Engineering Magazine: Fall 2017

CARNEGIE INSTITUTE OF TECHNOLOGY ALUMNI MAGAZINE // FALL 2017 // ENGINEERING.CMU.EDU

CARNEGIE MELLON ENGINEERING

JOAO DIOGO DE MENEZES FALCAO PH.D. CANDIDATE, ECE CARNEGIE MELLON UNIVERSITY SILICON VALLEY

From the Dean

Contents FALL 2017

Entrepreneurship is part of the College of Engineering’s DNA. This includes how we create and implement new initiatives as a College, how we teach, how we do research, and the kind of research we do. This also includes the outlook, goals, and impact of our graduates. Because of our foundation in entrepreneurship and our capacity to leverage our world-class technology and innovation education in Pittsburgh with our forward-looking and rapid approach to entrepreneurship in Silicon Valley, we created the M.S. in Technology Ventures. This new professional degree teaches how to create technology-based businesses, be they startups or ventures within established companies.

The Master of Science in Technology Ventures (MSTV) program, launched in January 2017, teaches engineers how to thrive in the dynamic technology environment as both innovators and entrepreneurs. Students enhance

their

engineering

skills

while

exploring the ins and outs of launching ventures through coursework on finances, legal frameworks, and agile marketing. The MSTV program is multi-disciplined in that students from all of the College’s engineering programs have skills to offer. The program is team-driven, and students with varied backgrounds generate new and rich perspectives for solving problems.

Students are excited about the program because they gain handson experience in the startup ecosystem of Silicon Valley. Through

03 FEATURE 03 Radio City 07 The Muscle Suit 08 Website Evaluates Open Source Software 09 Drones in the Matrix 10 Ian Lane Puts the Smart in Smart Car 11 Decision Systems That Respect Privacy and Fairness

07 RESEARCH 13 Tracking Firefighters Through Heat and Smoke 15 Sensing Danger in Severe Environments 16 Deceptively Clean Vehicle Emissions 17 Innovation in Brain Imaging 18 An Easier Prescription

19 DEPARTMENT NEWS

internships, they gain exposure to emerging concepts and learn how to take innovations to market. Amplifying this notion, the MSTV program links with the Venture Bridge program so students can

21 INSIDE THE COLLEGE

launch their own ventures by tapping the resources provided by

21 NSF CAREER AWARDS

the SV campus, alums, angel investors, and venture capital firms.

23 INI Director Receives Professorship

The achievements that students attain in all of our Silicon Valley

in Information Networking

programs are due in part to the relationships they build. CMU-SV

23 Carnegie's Medallion

students have access to well-connected professors, industry leaders,

24 PPP: The Strongest Team

and active Bay Area alumni who provide internships, mentoring, and

in DefCon History

funding. Yet, these relationships are not one-sided—our students

25 Innovator Leads Energy Institute

give back. Alumni and industry leaders say that they value and benefit

27 Honoring Angel Jordan

from our students’ knowledge and perspectives on problem-solving.

Featured in this issue are some of the robust research projects underway at CMU-SV. The faculty there are delving into topics ranging from network development for smart cities to advanced human-computer interaction systems for cars, and with each of the many research projects we undertake comes the opportunity to do what we do best: educate new engineers.

Sincerely, James H. Garrett Jr.

29 STUDENT NEWS 29 A Handheld Device to Monitor for Breast Cancer 31 Self-Driving Buggies Make History 32 INI Students Innovate at Internships

33 ALUMNI 33 Father Like Son

CELEBRATING 50 YEARS

as Carnegie Mellon University during Homecoming Weekend November 10–11, 2017 | Carnegie Mellon University | Pittsburgh, PA

We want to hear from you Send email to stokes@cmu.edu Please include your name and, if applicable, major

EDITOR Sherry Stokes (DC’07) DESIGNER

Tim Kelly (A’05, HNZ’14) COPY EDITOR

Emily Durham

WRITERS Krista Burns Jessica Corry Hannah Diorio-Toth (DC’17) Adam Dove Emily Durham

Alexandra George Samantha Jamison Lisa Kulick Vidya Palepu Daniel Tkacik (E’13)

and date of graduation.

ENGINEERING FALL 2017

02

SILICON VALLEY

RADIO CITY

IN SILICON VALLEY

If you ask Bob Iannucci to describe a radio, he is likely to take his cell

WHAT’S RADIO GOT TO DO WITH IT?

phone out of his pocket and hold it up to you.

“When you look at this, you see a cell phone,” he says. “When I look at it, I see 8 antenna and 11 radios.”

Iannucci is the director of the CyLab Mobility Research Center

In essence, a radio is just a means for remote communication. It’s a device that can transmit and receive electromagnetic waves of radio frequency that carry information, such as audio and data. The

at Carnegie Mellon University, and resident at CMU’s Silicon Valley

information is captured in these radio waves by altering the waves’

campus. His cell phone exemplifies how prevalent radio technology is

amplitude, frequency, and phase (a process called modulation). The

in our daily lives. If it weren’t for radios, we wouldn’t type on Bluetooth

waves are then transmitted through the air to an electrical conductor,

keyboards, wouldn’t engage in wireless video chats, wouldn’t be guided

where they are demodulated to recover the original information.

safely to our destinations by GPS mapping. These technologies illustrate what radios do for us now. But what will they do for us in the future?

This is how music can travel from a radio station’s antenna to your car’s stereo, or how your voice travels from one cell phone to another, and even how GPS data is extracted from satellites circling the earth.

Bluetooth is radio. Your Wifi and LTE receivers are radios, too. But according to Iannucci, the future of radio technology will take us far beyond these applications to a whole new computing platform that will be just as revolutionary as the smart phone.

ENGINEERING FALL 2017

04

“We’ve created an experimental, mostly software-defined

Once again, Carnegie Mellon is leading the charge toward the

testbed for advanced wireless research, called CROSSMobile,” says

next generation platform—this time, an open, intelligent, flexible,

Iannucci. “We’re using it to study critical elements that will make

spectrum-collaborative communication network that merges

up the next-generation communication and computing platform.

communication and computation all the way to the smallest sensor

We start with today’s wireless networks and study how they can be

devices. Carnegie Mellon seeks to make this research relevant and

improved. But rather than just considering how to stream videos

applicable by conducting studies of the key problems in a real-

faster or how to improve person-to-person messaging, we are

world context. Integrating the CROSSMobile project with CMU’s

converging on a vision of how communications and computation

Crowdsourced Smart City project weaves together low-power, in-

will come together to create a wholly new computing platform.”

the-city sensing and embedded computing with the network itself.

Iannucci’s platform is based on the idea that communication—

In essence, Iannucci and his team imagine that future networks,

not only between people, but between devices—is the key to

together with their associated sensing devices, will emerge as the next

next-generation computing technologies. Platforms of the past

major computing platform. But bringing this vision to reality requires

– mainframes, minicomputers, and PCs – largely focused on

solving a programming problem. To the programmer, the cloud,

computation. But the mobile revolution brought about a shift in

the network, and the devices together must be made to look like a

thinking. Computation and communication became equals. Looking

coherent ensemble rather than so many separate yet interconnected

ahead, however, we may come to see communication dominate

subsystems. Iannucci calls this concept “edgeless computing.”

computation in emerging platforms. Consider the so-called smart city.

From machine sensing and predicting traffic patterns, to generating

CHANGING THE NATURE OF NETWORKS

wireless emergency alerts and aerial drone surveying; smart cities will be integrated with millions of devices, talking back and forth

“If we think about smart cities and say, smart intersections, as cars

constantly on so many radios while aggregating and processing shreds

come into the intersection, we want make sure that they don’t collide

of information to generate insights. And at the center of it all will be

or hit pedestrians,” says Iannucci. “The sensors in the intersection

the next-generation wireless network.

at that time—in cars, in the pavement, in the signaling systems,

“Together with my colleagues ECE Associate Professor Anthony

and in the smartphones of pedestrians near the intersection—can

Rowe and Assistant Professor Swarun Kumar, we’re putting traffic

be federated, and what they tell us at that instant can help avoid

sensors into the city environment, and these sensors need robust,

accidents and improve efficiency; but only if the wireless network

low-power wireless communication capabilities,” says Iannucci.

does its job right.”

“We’ve built mechanisms to allow these sensors to self-adapt to the

Connecting sensed data—such as where the cars are on the

challenging radio conditions of the smart city – it’s the ‘can you hear

road—with time on a millisecond scale relies on mechanisms

me now’ question for billions of city-embedded Internet of Things

that don’t exist in today’s networks and sensor devices. Adding

devices. Taking a cue from Carnegie Mellon’s past, we are approaching

the mechanisms is possible, but effectively coordinating them is

the understanding of networks for smart cities from a systems point of

an enormous challenge for the programmer. Currently, Iannucci

view. What’s emerging is the need to stop thinking about networks and

is working on new approaches that would relieve many of these

start thinking about platforms. This work brings together past research

complexities. Making the creation of smart city ‘apps’ about as easy as

in distributed systems, real-time computation, wireless networking

creating smartphone apps will be essential to bring about the same

and low-power embedded computing into a new model that seeks to

level of societal and economic impact the smartphone has had. This

radically simplify the programming task for such an ensemble.”

is what Iannucci is talking about when he talks about a new kind of

This initial traffic sensing research is funded by the University Transportation Center’s Technologies for Safe and Efficient

computing platform.

In the smart intersection, the lower the network’s latency—or

Transportation (TSET) Initiative, and Iannucci is currently working

the time it takes for data to be transported over a network—the

closely with faculty at the Pittsburgh campus to deploy this sensing

fewer accidents will occur. With current technological capacity, all

technology around the city.

communication networks go through the cloud. If a smart traffic light must turn red because two cars are about to collide, a signal would

BROADBAND AND THE CMU NETWORK

have to go from the light’s sensor, up through the cloud, then back down to the light. This could take up to 2000 milliseconds. To prevent

Today, we take the power of parallel computing for granted. Google, Amazon, and the rest of the tech world’s biggest companies use

an accident, the latency needs to be more like 2 milliseconds.

“We are trying to change the nature of networks of IoT devices that

parallel computing platforms, or huge data centers filled with

will allow programmers to create apps that, at the same time, minimize

computers that are linked together to provide seemingly unlimited

latency, achieve time synchronization and minimize power—especially

storage and computing power. But as little as 30 years ago, some

important for install-and-forget smart city devices. Our work goes beyond

of the first, groundbreaking work in high-performance parallel

the so-called edge of the network to create an integrated system for

computing was done here at Carnegie Mellon. The C.mmp and CM*

apps that encompasses cloud, network and device: edgeless computing.”

systems projects provided proof that computers could be harnessed

Iannucci says. “Edgelessness – breaking down the assumption that the

in parallel to do large, complex tasks, leading to the parallel data

network has, or should have, an edge is essential to programmable, low-

centers that now power much of the world around us.

latency distributed computing in the smart city and beyond.”

Because of the large amount of data that is constantly

Optimizing the 21 century radio network st

moving back and forth across communication networks, latency optimization isn’t the only barrier.

The sheer wealth of data can overload a network, rendering it unusable. Many of us routinely experience this—you look down at your phone and it appears you have service, yet you can’t make a call or access the internet. Midnight rolls around on New Year’s Eve and your texts won’t go through, thanks to the vast number of texts being sent at that exact moment.

CMU-SV Professor Carlee Joe-Wong is interested in how to create networks that can stand up to this high traffic. Using models of pricing and economic incentive, Joe-Wong is researching how to get users to shift their cellular usage from one interface to another.

“If users experience a lack of service on their cellular networks, they will switch over from the cellular radio to the wi-fi radio, thereby decongesting the cellular network. Now imagine that principle being extended to Bluetooth or Femtocell or other radio technologies being developed,” says Joe-Wong.

In this model, users have to experience disruption before manually switching over to a different radio.

But Joe-Wong envisions a network where users switch seamlessly between radios so that network congestion can be completely eliminated, and everything flows smoothly with no dip in coverage.

She hopes to achieve this through monetary incentive by offering users discounts to switch between radios at high-traffic times.

One of the difficulties is that these networks are owned by different companies, so you can’t dictate what price you're going to charge at what time. But certain providers are starting to pool their networks together: Google has an initiative plan called Google-Fi where users can switch between Sprint, US Cellular, and T-Mobile networks.

“Our research looks at the efficiency and the economics of that,” says Joe-Wong. “How much does Google have to pay these providers in order to get them to like partner with Google? How much can Google charge users in order to get them to subscribe? And then does that all work out so that Google as a virtual service provider can make a profit? These are the types of questions we will need to answer to make these kinds of networks a viable option in the future.”

When it comes to cellular networks, overloads can be an inconvenience and lead to customer dissatisfaction. But when it comes to smart cities technologies such as smart streetlights that regulate traffic, a downed network can be a matter of life or death. Large-scale elimination of network congestion will go a long way toward providing communities with the network reliability necessary to ensure public safety.

ENGINEERING FALL 2017

06

THE MUSCLE SUIT SILICON VALLEY

For centuries, professional athletes have been doing everything they can to get ahead. From legal means, like hiring the best trainers and purchasing the best equipment, to less legal means, like pharmaceutical enhancement, athletic advantages come in all forms.

But thanks to CMU-SV Professor Pei Zhang, the modern athlete now has access to the newest advancement in performance enhancing technology: data.

Though Zhang’s vibration sensors were initially developed to track individuals’ movements through a building, he quickly realized they were sensitive enough to measure smaller movements, and even movements inside the body. By shrinking these sensors down and sewing them into a bodysuit, Zhang has created the Muscle Activity Recognition System (MARS).

Developed in collaboration with Frank Mokaya of CMU-SV and Cynthia Kuo, Quinn Jacobson, and Brian Nguyen of Vibrado Technologies, MARS employs real-time motion capture and visualization in any lighting. Using these vibration sensors, the suit is able to measure the micro-movements in each of the wearer’s muscles.

“Using this body sensor suit to measure the fine grain vibrations of your body we can find out which muscle you are activating, how hard you are activating these particular muscles, and how tired these muscles are," says Zhang.

Inside the suit are 18 sensor nodes with accelerometers, gyroscopes, and magnetometers. Using these vibration sensors, the suit maps the activation of particular muscle groups onto the on-screen avatar. When the muscles are first activated, the avatar highlights them in green, but as they fatigue over the duration of the exercise, they change to orange, and eventually red.

This level of data has never before been available to athletes and trainers, and opens up a whole new realm of possibility. By allowing athletes to monitor their muscle fatigue in real time, MARS gives them the ability to correct their posture and movements as they train.

Practicing proper movements can greatly reduce the chance of injury, and help athletes to learn skills more quickly and safely.

“This technology can enable both professional and amateur athletes to accurately track the extent of their exercise,” says Zhang, “in order to push themselves to their limits, but not over.”

With Zhang’s vibration sensor suit, athletes of the future will be able to step up their game with the most powerful performance enhancer of all—knowledge.

Website OSSpal.org Evaluates Open Source Software The spirit of contribution and cooperation is alive and well in the world of programming software. Witness, for example, the continued growth of open source software. With this booming popularity, CMUSV Professor Tony Wasserman has made his own contribution with OSSpal.org, a website that provides evaluations and ratings of open source software.

With nearly three decades of open source experience, Wasserman saw all of this open source software becoming available with no way for users to determine which software was worthwhile. How does the software perform? What are the best open source options for specific application needs? Which one will do what I need it to, and which ones won’t?

These questions are answered with OSSpal’s database of software evaluations and ratings. In the spirit of contribution within the open source world, these evaluations and ratings are conducted by the site’s users.

Once registered, users are able to browse or search by name or category. For each included project, the evaluation includes some general aspects (installability, usability, robustness, security, and scalability) as well as other aspects specific to its software category.

Those ratings can then be viewed by other users searching for software to use for their own projects.

“If you want to experiment or evaluate open source software,” Wasserman says, “we’ve done some of the work for you.”

What makes open source software appealing is its accessibility.

Unlike the more traditional and well-known proprietary software, which requires the purchase of a license and which has a source code that is inaccessible to users, open source software is so-named because its source code is open for anyone to access, inspect, modify, and enhance.

This ease of access, ease of modification, and lack of cost makes open source software extremely appealing to start-up companies.

“OSSpal is a way to help people find high-quality open source software that they can consider for use in their own project,” he says.

According to Wasserman, there is no other site acting as a comprehensive, user-focused resource for projects, allowing users to learn what a software does, and if it does it well.

However, open source software does present its own set of concerns. From fears about security and functionality, to questions about technical support, it has taken time for the idea of an open source-code platform to become widely accepted and widely used. But now, says Wasserman, it’s everywhere.

ENGINEERING FALL 2017

08

Drones in the Matrix

In the 1999 film The Matrix, a young hacker

Iannucci and his team, consisting of

able to create a photorealistic, virtual

named Neo learns that the world as he

ECE researchers Ervin Teng, Joao Diogo

training environment, identical to the

knows it is a virtual simulation—and with this

de Menezes Falcao, and Cef Ramirez, are

real-world environment the drone will be

realization comes the ability to use this virtual

working on a number of projects to train

encountering,” says Iannucci. “By training a

world to his advantage. In one scene, Neo,

drones through real-time deep learning. The

virtual drone to identify the desired object in

who has no combat training whatsoever,

first, called SMILE, uses a camera mounted

the virtual world, then uploading that data

downloads an extensive knowledge of martial

on the drone to take pictures of the drone’s

to the real drone in the real world, the real

arts into his brain, making him a Kung Fu

current surroundings and send those images

drone will ‘remember’ everything the virtual

master in mere seconds.

back to an operator on the ground. If you

drone learned.”

“What we’re doing is just like the Matrix,”

want the drone to track down a particular

Because the training simulations are

says Bob Iannucci, distinguished service

red convertible, for instance, some of the

done in the virtual world, operators can

professor of ECE at CMU-SV. “We’re hooking

images it sends back will contain the desired

run them much faster than real-time. In the

into the visual cortex and downloading

car, and some will not. Using a computer

future, by networking multiple computers

information that would typically take much

interface, the operator lets the drone know

running multiple simulations together, the

longer to learn in a very short amount of

whether or not the object it’s looking for is

time reductions are potentially limitless.

time. Only, we’re teaching drones.”

in the image. After just a few minutes, the

And now, certain more dangerous tasks—

drone can identify the desired object in the

such as navigating collapsed buildings or

particular car. To teach the drone which car

Imagine you want your drone to track a

images with enough accuracy to track it

complicated industrial sites—can be done

to track, you can apply machine vision and

across distances.

virtually, so that operators don’t have to

deep learning techniques which work by

Just like Neo, however, drone operators

worry about expensive drones crashing

aggregating large amounts of photo data to

don’t always have the luxury of practicing in

help the drone identify an object in a variety of

the real world. Due to lack of time or lack of

“This technology is broadly applicable,”

situations. These algorithms take a significant

available data, training a drone to identify an

Iannucci says. “Not only can it track individual

wealth of data to train, however, and these

object in space or navigate an environment

cars or people, but it can be used for many

substantial training sets are hard to obtain.

in the real world can be challenging. It’s this

other things, such as inspection of industrial

difficulty that led the team to develop the

sites or pipelines. It may not teach them Kung

but you still need to tell the drone what it’s

Virtual Image Processing Environment for

Fu, but the variety of tasks that VIPER has

looking for, in real-time, using only the data it

Research, or VIPER.

the potential to teach drones to perform is

So what if you don’t have a lot of data,

can collect from its surroundings?

“Using a video game engine, we’re

before they’re properly prepared.

virtually limitless.”

Ian Lane Puts the Smart In Smart Car

One of the most popular tropes in sci-fi—be

Typically, in order to access GPS directions

deep learning systems, Capio is able to build

it James Bond, Knightrider, what have you—is

or find nearby restaurants, drivers rely on

upon its ability to pick out individual voices

the intelligent car sidekick. It can drive itself,

their smartphones to tell them where to go.

in order to get better and more accurate

see its surroundings, carry on conversations.

But this can present a number of problems.

through its interactions with people over

With advances in autonomous vehicle

Drivers looking down or typing on their

time—the same way that children learn to

and sensor technologies, engineers are

phones while driving is a serious safety

pick their parents’ voices out of a crowd.

every day bringing us closer to this science

concern, and one of the leading causes

fiction reality. But when it comes to certain

of fatal accidents in the U.S. Some newer

human-computer interaction system not

elements, such as intelligent navigation

vehicles have voice-activated GPS systems

only solves the problem of drivers using their

and conversation, the real world has fallen

installed directly into the car, but if there is

phones while on the road, but combined with

woefully behind. Until, that is, CMU-SV

any background noise in the car, these voice-

its onboard computer vision technology, it

professor Ian Lane created Capio.

activated systems break down.

allows users to use their hands in a variety

“We’re taking this technology to a level that

“Though these systems have improved

This hands-free, contextually aware,

of more helpful ways. If you’re driving down

mimics human capacity,” says Lane, “enabling

dramatically over the last few years—Siri

the road and see a restaurant that looks

people to interact with a machine in the same

and Alexa, for instance—there are still

interesting, all you have to do is point to it

way they would interact with a human.”

challenges,” says Lane. “Often, these systems

and ask Capio if it’s any good. Using GPS

Outwardly, Capio doesn’t look like

fail when there are many people speaking at

data and full internet connection, Capio

much—a simple black bar mounted on the

the same time. For instance, if you’re driving

can pinpoint the exact restaurant you are

car’s dash, equipped with a camera and audio

in your car and the kids are screaming in the

referring to and pull up its online reviews,

sensors. But despite its humble appearance,

back seat, the system doesn’t work. We’ve

then give recommendations for similar

Lane’s technology solves all of the problems

developed technology that understands

restaurants nearby that you might like as

of former systems. Using computer vision-

individual speakers, so even if there are

much, or more.

based approaches, the system can track

three or four people speaking at the same

the movements and gestures of every face

time, the system can pick out one person

for Capio,” Lane says. “The future of

in the car. This way, the car can follow the

from that speech and recognize them with

contextually aware, human-computer

conversation in the same way a human can,

high accuracy.”

interaction systems will mean that eventually,

telling the difference between when the

Second, not only can the system follow

“These car systems are just the beginning

every interaction we have with the machines

passengers are talking to each other, and

conversation like a human does, it can also

we encounter on a daily basis will feel just as

when they’re addressing Capio directly.

learn in the same way as a human. Using

comfortable as speaking to a person.”

ENGINEERING FALL 2017

10

Decision Systems That Respect Privacy And Fairness

Researchers seek to add a layer of humanity to artificially intelligent systems

Increasingly, decisions and actions

ECE at Carnegie Mellon’s Silicon Valley

Tschantz, senior researcher at the International

affecting people’s lives are determined by

campus, will lead a $3 million National

Computer Science Institute in Berkeley.

automated systems processing personal

Science Foundation project on accountable

data. Excitement about these systems has

decision systems that respect privacy and

automatically account for why an automated

been accompanied by serious concerns

fairness expectations. The project seeks

system with artificial intelligence components

about their opacity and threats they pose to

to make real-world, automated decision-

exhibits behavior that is problematic for

privacy, fairness, and other values. Examples

making systems accountable for privacy

privacy or fairness,” says Datta. “These

abound in real-world systems: Target’s use

and fairness by enabling them to detect

explanations then inform fixes to the system

of predicted pregnancy status for marketing;

and explain violations of these values. The

to avoid future violations.”

Google’s use of health-related search

project will explore applications in online

queries for targeted advertising; race being

advertising, healthcare, and criminal justice,

explanations for systems that employ

associated with automated predictions of

in collaboration with domain experts.

statistical machine learning,” adds

recidivism; gender affecting displayed job-

The CMU project team includes Matthew

“A key innovation of the project is to

“The hard part is creating such

Mengshoel. “But doing so is critical, since

related ads; race affecting displayed search

Fredrikson, assistant professor of computer

these methods are increasingly used to

ads; Boston’s Street Bump app focusing

science, and Ole Mengshoel, principal systems

power automated decision systems.”

pothole repair on affluent neighborhoods;

scientist in ECE. The project also marks a

Amazon’s same day delivery being

collaboration between CMU, Cornell Tech, and

fairness in decision systems, the team must

unavailable in black neighborhoods; and

the International Computer Science Institute.

first provide formal definitional frameworks

Facebook showing either “white” or “black”

Additional contributors are Helen Nissenbaum,

of what privacy and fairness truly entail.

movie trailers based upon “ethnic affiliation.”

professor of information science at Cornell,

These definitions must be enforceable

Thomas Ristenpart, associate professor of

and context-dependent, dealing with both

computer science at Cornell, and Michael C.

protected information itself—like race,

Recognizing these concerns, CyLab’s Anupam Datta, associate professor of

But in order to address privacy and

gender, or health information—as well as

property of data processors and compromise

proxies for that information, so that the full

the confidentiality of the training data

scope of risks is covered.

subjects, as demonstrated in the prior work

“Committing to philosophical rigor, the

of many on the team,” says Fredrikson.

project will integrate socially meaningful

“Unfortunately, we don’t yet understand

conceptions of privacy, fairness, and

what machine learning systems are leaking

accountability into its scientific efforts,”

about privacy-sensitive training data sets.

comments Nissenbaum, “thereby ensuring its

This project will be a great opportunity to

relevance to fundamental societal challenges.”

investigate the extent to which having access

“Although science cannot decide moral

to prediction functions or their parameters

questions, given a standard from ethics,

reveals sensitive information, and, in turn,

science can shed light on how to enforce it,

how to improve machine learning to be more

its consequences, and how it compares to

privacy-friendly.”

other standards,” says Tschantz.

Another fundamental challenge the team

Datta has assembled a truly interdisciplinary team of researchers

faces is in enabling accountability while

for the project. Combining the skills of

simultaneously protecting the system owners’

experts in philosophy, ethics, machine

intellectual property, and the privacy of the

learning, security, and privacy, Datta hopes

system’s users.

to successfully enable accountability

“Since accountability mechanisms require

in automated decision systems—an

some level of access to the system, they can,

achievement that would add a layer of

unless carefully designed, leak the intellectual

humanity to artificially intelligent systems.

ENGINEERING FALL 2017

12

ENGINEERING.RESEARCH

TRACKING FIREFIGHTERS THROUGH HEAT AND SMOKE

Tracking firefighters in burning buildings

The researchers intend to outfit the

is fraught with challenges. Smoke renders

air packs that firefighters wear with

laser- and vision-based tracking technologies

inexpensive Smart RF Tags to communicate

useless, while heat and flames will obliterate

their location to a small number of beacons

pre-installed monitoring devices. GPS isn’t

strategically deployed around the fire.

an option either, because it doesn’t work

Signals will relay from the firefighters to the

indoors. Another constraint is the need

beacons. This data will then stream to the

for speed—when firefighters arrive on the

safety chief, who will monitor on a laptop

scene, they don’t have time to operate

what the firefighters are facing.

complex technology.

The outdoor beacons would be

What emergency responders need is a

outfitted with a GPS receiver, radios, and

system that tracks firefighters inside blazing

sensors. These technologies will allow for

structures and that will operate without prior

communications deep within the building

knowledge of a building’s floor plan.

and help estimate a firefighter’s elevation.

ECE professors Anthony Rowe and

For example, are the firefighters on the 2nd

Bruno Sinopoli, both pioneers in the area

or 10th floor? The firefighter’s wearable

of indoor positioning for over a decade,

units, which would cost under $100, would be

are now advancing wireless broadband

outfitted with similar technology, only the GPS

communications for use in extremely hostile

would be replaced with an IMU to measure

environments such as burning buildings.

their movement and orientation over time.

The beacons situated at the point of

We’re developing a positioning system to locate firefighters inside burning structures By combining three emerging

ingress into the building will mark the firefighters’ starting location. Once inside the building, the firefighters’ air pack devices will send radio signals to convey where the teams are in relationship to the beacon and each other. Long-range radios will send data from inside the building to the beacons.

Developing a mobile indoor positioning system that operates on the fly is difficult enough without the added challenges of

technologies, the researchers are creating a

heat, smoke, and flames, but Rowe and

failsafe system to reveal where firefighters

Sinopoli have significant experience in

are in a building along with their orientation,

this area. In 2015, their students won the

or the direction they are facing. Orientation

infrastructure-based 2015 Microsoft Indoor

tracking is critical so that firefighters working

Localization Competition. They have also

in zero-visibility situations can be directed

deployed a sophisticated positioning system

(via radio communications) where to turn or

inside the David Lawrence Pittsburgh

when to enter or exit an area.

Convention Center, which spans 1.5 million

To guide their work, the researchers met with firefighters to learn their operational

square feet.

Building off these successes, the team

procedures. When firefighters arrive at

has the confidence this new project will

a fire, they are dispatched in teams of

result in a prototype that will transition to the

two, and a safety chief checks in with

commercial market. Rowe and Sinopoli work

the teams every 15 minutes by radio. If

closely with Bosch's Research and Technology

communications are lost, the safety chief

Center in Pittsburgh, which has keen interest

sends in an intervention team to find the

in indoor positioning systems. Other product

missing firefighters. This can become

development opportunities exist too, with

exceedingly dangerous.

the creation of air pack units for firefighters

On the technical front, the team analyzed various positioning technologies, and determined it will take a combination of

and visualization tools for sturdy laptops or tablets for safety personnel.

The U.S. Commerce Department’s

devices—inertial measurement units (IMU),

National Institute of Standards and

ultra-wideband radios (UWB), and long-range

Technology (NIST) awarded Rowe and

wide-area network radios (LR-WAN) — to

Sinopoli a three-year grant for $782,000 in

pinpoint firefighters.

2017 for this project.

ENGINEERING FALL 2017

14

MSE alumni Jason Gu and Jacob Melby and MSE Professor Bob Davis have developed a hydrogen sensor that can prevent catastrophic explosions in severe environments. They, together with MSE Professor Lisa Porter, founded SenSevere LLC to produce hydrogen sensors for harsh environments, such as chlorine production cells and nuclear reactors.

During the production of chlorine, high concentrations of hydrogen

Peering Inside a Nuclear Reactor

are produced and must be swept out of the system. If hydrogen is accidently allowed to mix and react with chlorine, the reaction produces hydrochloric acid (HCl), which, in turn, produces so much heat that it often causes massive explosions.

The hydrogen sensor, which was developed at Carnegie Mellon, is vital in materials science because detecting hydrogen during chlorine production can prevent explosions and loss of life. SenSevere’s device is unique because it can go inside working chlorine production cells and endure the harsh chemical environment created by chlorine.

Before this device, no other sensor on the market could withstand the corrosive environments generated during chlorine production.

“The importance of the sensor has to do with the robustness of the sensor itself,” said Gu. “Basically, the more chemically stable the materials that you’re using in your sensor, the less it will interact with its environment. The sensor enables us to be more flexible in terms of the things that we choose for it to interact with, because it itself is very stable.”

Besides chlorine production, Davis explained that the sensor can be used in nuclear reactors. Hydrogen buildup in nuclear reactors typically led to massive explosions because of high concentrations of hydrogen interacting with oxygen or water. Davis and Gu emphasize the importance of the SenSevere’s hydrogen sensor because it can provide sensing in environments that could never be sensed before.

“If you think about the inside of a nuclear reactor,” said Gu, “the further in you get, the less information you have about what’s going on, and the slower that information comes. If you’re at the very outside of the core, you can bring a handheld sensor in. But as you go further and further into the core, the information gets much more delayed and less accurate because you can’t send a person in, and other sensors don’t work as well in there.”

Sensing Danger In Severe Environments New sensor detects hydrogen to prevent explosions

“The hydrogen sensor provides us with faster and more accurate information about some of these dangerous environments and processes that are present.”

The inspiration for the company’s device came from the team’s interest in Group-III nitride films and devices. For this project, they tested an aluminum gallium nitride (AlGaN) high-electron-mobility transistor (HEMT) to evaluate its capacity as a sensor material. Gu thought the transistor might be sensitive to hydrogen, so the team sought to test his hypothesis.

“We set up a laboratory in Wean Hall, and Gu and Melby began testing the transistor for hydrogen sensitivity at various concentrations, temperatures, and pressures,” said Davis. “And lo and behold, it turned out to be a really excellent sensor for hydrogen.”

“This is where knowledge, understanding, and a little luck helped to create a device which could be brought to commercial fruition.”

Deceptively Clean Vehicle Emissions

23 million Americans live in areas that violate the federal standard for fine particulate matter. Bringing these areas into compliance will likely require addressing secondary organic aerosol pollution.

For years, automakers have been working to reduce pollutant levels coming out of motor vehicles’ tailpipes. Airborne particulate matter, for example, is present in automobile exhaust and has been shown to contribute to tens of thousands of premature deaths in the United States each year.

Researchers in the College of Engineering found that newer gasoline vehicles emit less particulate matter, but vapors in the “cleaner” exhaust form particulate matter in the atmosphere long after exiting the tailpipe. These secondary particles, or secondary organic aerosol, can be just as harmful to human health. The findings were published in Proceedings of the National Academy of Sciences.

MechE Department Head Allen Robinson and his research team

“Overall, we found that new and stricter regulations for gasoline vehicle tailpipe emissions will not be that effective at reducing human

investigated the formation of secondary organic aerosol (SOA) from the

exposure to secondary organic aerosol because of changing NOx

exhaust of gasoline vehicles in urban areas and the effects that more

levels,” says Robinson. “This feedback illustrates the complex coupling

stringent vehicle standards could have on SOA formation. Secondary

between different pollutants, which must be accounted for in models

organic aerosol is a major component of atmospheric fine particles,

used to develop control strategies.”

which negatively affect the human body and the earth’s climate.

Robinson and his colleagues found that emissions from newer vehicles meeting more stringent emissions standards formed less SOA than older vehicles, which means that replacing older vehicles

The team’s research ultimately highlights the importance of addressing SOA pollution in cities all over the country, from Los Angeles to Pittsburgh.

“About 23 million Americans live in areas that violate the current

with newer vehicles will reduce pollutants in the air. However, the

federal standard for fine particulate matter,” says Robinson. “SOA is

team also discovered that the formation of SOA strongly depends on

a major component of those particles, and bringing these areas into

nitrous oxides (NOx), which are abundant in cities and other areas

compliance—so that all Americans breathe cleaner, safer air—will

with high motor vehicle traffic.

likely require addressing SOA pollution.”

ENGINEERING FALL 2017

16

Innovation in Brain Imaging

Writers and scientists have searched for an apt technological

modality currently has, such as high spatiotemporal imaging

analogy for the human brain, often comparing it to a

resolution, while still being portable. It is also more than ten

computer. For Pulkit Grover, assistant professor of ECE and

times cheaper than other imaging technologies such as MRI or

the Center for Neural Basis of Cognition, this analogy couldn’t

Magnetoencephalography (MEG), which is important for doctors

be more fitting. Grover and his team explore how information

who will use this for treating and monitoring their patients.”

flows through computer networks (e.g., coding systems,

Grover and ECE Ph.D. student Praveen Venkatesh

cyber-physical systems, and low-power wireless systems),

established the first-ever fundamental limits on EEG imaging

and they apply these information theory principles to brain

and showed that the reason that most neuroscientists believe

imaging systems. This cross-disciplinary research approach

EEG inherently has low resolution is incorrect. These limits

bridges mathematical theory with clinical applications—

show how an earlier study, which suggested that low-density

striving to improve the treatment of neurological disorders

systems (with a hundred or so electrodes) obtain the best

such as epilepsy.

possible imaging, was misunderstood at-large in the field.

“It is exciting to apply my research in the neuroscience and

Grover and Venkatesh explored the question of how many

neuroengineering space because I am tackling information

electrodes should be used to obtain the best imaging results.

theory challenges that have the potential to impact the quality of

“If you improve both the data analysis and the number

life of patients, or make a doctor’s diagnosis faster and easier—

of electrodes for EEG systems, then you can improve the

and that is the goal I’m always working toward,” says Grover.

resolution dramatically,” says Grover.

Grover and his team have been focused on improving

The study, titled “An information-theoretic view of EEG

the resolution of EEG neural imaging technology, a portable

sensing,” was published in the Proceedings of the IEEE.

and non-invasive brain imaging system. Their research

The research was conducted as part of Carnegie Mellon’s

goes against the trend in the field of neuroscience. Many

BrainHub. Grover and Venkatesh are collaborating across

researchers believe that EEG systems are fundamentally

the university to validate their fundamental results and bring

limited to imaging resolutions that are too low to be effective,

them into practical systems. They are also working with Mark

and that it is impossible to improve the resolution of these

Richardson, an epilepsy neurosurgeon at the University

systems beyond their current levels.

of Pittsburgh, to obtain clinical validation and establish

“The current overarching view in both clinical and neuroscientific communities is that a different imaging modality

relevance in epilepsy.

“To change the widespread perception of EEG technology

needs to be invented. However, it is our view that the potential

and get these systems into clinical practice, we need more

of EEG has been severely underestimated,” explains Grover.

experimental validation of this theory,” Grover concludes.

“We are working towards building the first ‘Ultra-Resolution

“We are well on our way to getting these validations, and I’m

EEG’ platform. This platform offers benefits that no other

looking forward to what the future holds for this research.”

An Easier Prescription

Type I diabetes patients typically inject insulin

with colleague Alan Russell, professor and

moving the model protein through the

several times a day, a process that reduces

director of the Disruptive Health Technology

intestinal wall.”

quality of life. Injectable medications are

Institute. With Whitehead’s background in

also associated with noncompliance, which

drug delivery systems and Russell’s expertise

identified a chemical structure called

can result in complications for patients with

in polymer-based protein engineering, the

phenylpiperazine that increases the

chronic disease and dramatic increases in

team developed a novel solution. Their

permeability of the intestine. By surrounding

healthcare costs.

research was published in "The Journal of

the protein in a polymer made from

Controlled Release".

phenylpiperazine, the model protein easily

So, what’s holding doctors back from prescribing a simpler solution, like an insulin

Using a technique called Atom Transfer

To tackle this challenge, Whitehead

passed across intestinal cell barriers.

pill? Our own digestion systems are to

Radical Polymerization (ATRP), which was

Notably, the research team increased the

blame—because in this case, they function

developed at Carnegie Mellon by chemistry

transport of the model protein without

too well for their own good.

professor Krzysztof Matyjaszewski, the team

increasing the transport of other harmful

created a packaged protein that survives

compounds, such as waste products, across

ingest as food, even if this is a therapeutic

digestion-like conditions and is easily

the intestine.

protein drug like insulin. Proteins that enter

transported across the intestinal barrier in

the stomach are digested into individual

a cell culture model. The protein used in

because we have demonstrated that

amino acids and lose any intended

this study served as a model for therapeutic

polymer conjugation can be used to achieve

therapeutic function,” explains Katie

drugs such as insulin.

oral protein delivery. These findings give

“Our bodies see any proteins that we

Whitehead, assistant professor of ChemE. Even if the drug was able to make the

ATRP allowed Russell to attach a polymer

“We are excited about this research

rise to many more questions that we look

to the model protein. Once attached, this

forward to tackling, such as how the polymer

journey to the small intestine without being

polymer acted as a shield against digestive

structure and architecture influence the

digested, our body still wouldn’t be able

enzymes in the stomach. “Our team had

delivery process as well as whether these

to absorb it. Large protein drugs do not

developed a polymer that was very stable,

results translate in vivo,” says Whitehead.

permeate the intestinal lining, which means

enough that it would survive in hydrochloric

it is impossible for the drug to move into the

acid,” says Russell. “We felt confident that

to be an important introductory step in

bloodstream and start to work in the body.

we could use this polymer to protect the

their research towards developing oral drug

model protein from the stomach, but the

delivery systems that can be tested and

second challenge remained of selectively

used clinically.

Whitehead saw this drug delivery challenge as an opportunity to join forces

This project is thought by the researchers

ENGINEERING FALL 2017

18

DEPARTMENT NEWS BIOMEDICAL ENGINEERING BIOMEDICAL ENGINEERING Starting on February 1, 2018, Bin He will succeed Yu-li Wang as the next Head of the Department of BME. Bin He is currently the Distinguished McKnight University Professor of BME and the Medtronic-Bakken Endowed Chair for Engineering in Medicine at the University of Minnesota, and has earned many awards, including the IEEE Technical Field Award in BME and the Academic Career Achievement Award and Distinguished Service Award from the IEEE Engineering in Medicine and Biology Society.

“He has made significant contributions to biomedical imaging… and to brain-computer interface technology,” says Dean James H. Garrett Jr. “His pioneering work has transformed electroencephalography into an important dynamic, three-dimensional neuroimaging modality for noninvasive brain research and management of brain disorders.”

CHEMICAL ENGINEERING ENGINEERING

CIVIL & ENVIRONMENTAL ENGINEERING

In 2011, the Dow Chemical Company

In March 2017, Mitsubishi Hitachi Power Systems (MHPS)

launched a program called the Universi-

Americas President & CEO Paul Browning unveiled the initial

ty Partnership Initiative (UPI) to advance

results of a CMU index measuring carbon dioxide emis-

research in process development, energy,

sions from the U.S. electrical power generation sector. The

transportation, and consumer applications.

Carnegie Mellon Power Sector Carbon Index, developed by

As part of this program, Dow invested in

researchers including CEE Professor Costa Samaras and EPP

CMU’s Department of ChemE. Thus began

Professor Inês Azevedo, provides a comprehensive picture of

the five-year project funded by Dow that

the environmental impact of electricity production in the U.S

supported nine graduate ChemE students

and will measure the environmental impact of the U.S. power

from 2011 to 2016, allowing them to gain an

grid during the previous 12 months and over an extended

industrial perspective on their research.

period back to 1990. The CMU index will also provide a sum-

Braulio Brunaud and Sreekanth Rajagopalan, two ChemE Ph.D. students working in supply chain optimization under Dow’s UPI,

mary of how much electricity generation is from coal, natural gas, nuclear, and renewables.

The Index’s initial findings showed that U.S. power

currently work directly with Dow engineers,

producers had cut carbon dioxide emissions intensity by 24

ensuring a two-way exchange of ideas as well

percent since 2005. In June 2017, CMU and MHPS announced

as an opportunity for the students to develop

the release of the first quarterly Power Sector Carbon Index

industry contacts.

measurement update of the carbon dioxide emissions

This year, as the project came to an end,

intensity from the U.S. electrical power generation sector.

Dow has taken steps to continue interactions

In comparing the first quarter of 2017 to the first quarter of

between industry and research at CMU. Dow

2016, the index found that the U.S. power plant emissions

has chosen CMU as one of the very first uni-

averaged 955 lbs. of carbon dioxide per megawatt hour (CO2/

versities with which to renew the UPI.

MWh) in the first quarter of 2017, which was up less than 1 percent from the same time frame in 2016.

MATERIALS SCIENCE & ENGINEERING ELECTRICAL & COMPUTER ENGINEERING

MECHANICAL ENGINEERING MechE/EPP ENGINEERING & PUBLIC POLICY

Carnegie Mellon University’s Cleotilde (Coty) Gon-

If you want to do sensing or communications without

zalez, Christian Lebiere and Lujo Bauer are part

wires—think a cell phone or a radar system—you

of a team that has received a $6.2 million Multi-

need something called spectrum, which is, according

disciplinary University Research Initiative (MURI)

to EPP Professor Jon Peha, “a range of frequencies of

grant from the Department of Defense to prevent

electromagnetic waves.” However, the United States is

cyberattacks. The project, “Realizing Cyber Inception:

currently experiencing a spectrum shortage, in which

Towards a Science of Personalized Deception for

wireless service providers or companies with new prod-

Cyber Defense,” will develop deception tactics based

ucts can’t access or afford the spectrum they need.

on theories from cognitive science, computational

Peha is working to enhance wireless service by

game theory and computer systems engineering.

ending this spectrum scarcity. According to Peha,

These new tactics are expected to leap ahead of

fundamental technical problems are not causing the

attackers by moving towards active defense, where

spectrum shortage, but rather the ways we limit access

new cyber environments will make it impossible for

to spectrum, and the ways we protect systems from

attackers to determine what is real and what is de-

interfering with each other—methods that have hardly

ceptive. This new approach to cybersecurity is called

changed since the 1920s. According to Peha, we need a

“Cyber Inception.”

paradigm shift—and Peha is approaching this problem

Bauer, an associate professor of ECE, will collaborate with colleagues from the University of North

from several angles to develop this new paradigm.

“For 80 years, people developing wireless systems

Carolina at Chapel Hill and North Carolina State

really didn’t have to think about this,” he says. “But

University. “We want to be able to trap the adversary

we are in a new era, or we are entering a new era,

in a system that looks entirely real, but is in fact engi-

where thinking about spectrum efficiency will really be

neered to let us observe the adversary’s behavior,”

fundamental to anyone involved in wireless products,

says Bauer.

wireless service, and wireless technology.”

MATERIALS SCIENCE & ENGINEERING Researchers from MSE and MechE recently won the Pennsylvania State University Direct Digital Deposition (CIMP-3D) Modeling Challenge for Additive Manufacturing, sponsored by America Makes and the Defense Advanced Research Projects Agency (DARPA). The challenge’s objective was to identify the current accuracy of computational models for simulating the thermal and microstructural response of a material during the additive manufacturing process, with the goal of utilizing these tools as a means for accelerating the qualification of additive manufacturing for critical applications.

The team—which included Professors Chris Pistorius (MSE), Anthony Rollett (MSE), and Shi-Chune Yao (MechE), as well as Patcharapit Promoppatum (MechE Ph.D. student)—was able to accurately represent the complex motion of a heat source through the build of their material, and won the 1st place title.

MECHANICAL ENGINEERING The Internet of Things (IoT) is the term for everyday objects that now have internet capabilities: think smart watches, Fitbits, and smartphones. Engineers who work on mechanical systems now need to understand how mechanical and cyber systems interact and operate together. To help engineering students bridge the gap, MechE Professor Kenji Shimada and his co-instructor, Tomotake Furuhata, have debuted a new MechE course, “Robotic Systems and IoT.” The course provides graduate students and upper-level undergraduates with an overview of how robotic systems are integrated into a larger framework called the Internet of Robotic Things.

“The purpose of this class is to discuss the big trends in robotic systems and the internet of things and to provide two tangible examples: self-driving cars and factory robots,” says Shimada.

ENGINEERING FALL 2017

20

MATTEO POZZI

STEVEN CHASE

ASWIN SANKARANARAYANAN

NSF CAREER AWARDS The NSF Faculty Early Career Development (CAREER) Program is a foundation-wide initiative, offering prestigious awards to encourage faculty early in their careers to serve as role models in research and education, and to build the foundation for a lifetime of leadership in their field.

ENGINEERING.INSIDE THE COLLEGE

MECHANISMS OF SKILL LEARNING

link between any individual neuron and

enable vision systems to deal with a wider

movement is not known. To overcome this

range of objects and imaging conditions.

Most people can swing a hammer, but they

problem, Chase and his lab will use a brain-

Sankaranarayanan’s team will use this

can’t swing it with the speed and precision

computer interface, a device that allows the

research to build computational cameras that

of a master carpenter. The difference is

brain to control a computer cursor using

are better equipped to handle specific tasks

thousands of hours of practice and the

thought alone, and observe how neurons

than traditional cameras.

systematic organization of hundreds of

change when mastering control of the

thousands of the brain’s neurons.

device. By using a brain-computer interface,

INFRASTRUCTURE MANAGEMENT UNDER

Chase says the group can interpret how

UNCERTAINTY

“Practice makes perfect, in a pretty literal sense. When we improve at a skill over

changes in individual neurons combine to

time, it is presumably driven by coordinated

enable skill development.

changes in our brain’s neural representation

“Sometimes our brain actually requires us

Matteo Pozzi sees great potential in sensors and robotic technology that collect data

of how that movement should be completed,”

to rebuild a neural circuit in order to make

to help inform decision-making related to

explains Steven Chase, an associate

what was previously impossible, possible,”

infrastructure. The NSF sees great potential

professor of BME. “Yet, the link between how

Chase says. “With this award, we will go deep

in Pozzi, and has given the CEE associate

our brain reorganizes its neurons and how

into that process and answer the question:

professor a $500,000 CAREER Award to

we learn a new skill is still largely unknown.”

‘how do you rebuild those neural circuits?’”

suggest strategies that are optimal for

Chase was awarded a five-year, $800,000

collecting information and for taking

NSF CAREER Award to discover the link

STUDYING LIGHT TO ADVANCE COMPUTER

actions. Through integrating models and

between the neural reorganization and skill

VISION AND MORE

computational approaches, Pozzi hopes

learning. With the award, he will research the behavioral factors that drive skill learning.

Chase is a member of CMU’s BrainHub,

to optimize infrastructure operation and Aswin Sankaranarayanan, assistant

maintenance, and the continued collection

professor of ECE, received an NSF CAREER

of information.

which builds on its strengths in biology,

Award of approximately $532,000 to study

computer science, psychology, statistics,

how light interacts with materials using light

resources, data collection and this process

and engineering.

rays and their transformations.

of learning about the infrastructure must

An improved understanding of the

The project, titled “Plenoptic Signal

“Because we are managing such limited

be optimized,” he says, proposing that

science behind skill learning will have long-

Processing—A Framework for Sampling,

algorithms could offer guidance on where

term impact on the clinical understanding

Detection, and Estimation Using Plenoptic

and when to add more sensors, schedule

of the progression of various motor control

Functions,” will explore how light interacts

inspections, or conduct strategic testing.

disorders, such as Parkinson’s disease and

with objects in a scene by studying

stroke. His research may inform the design

characterizations of light that go beyond

of collecting information with the benefits

of targeted rehabilitation programs for

images. Specifically, Sankaranarayanan

of repairing various components, where

those patient groups.

will study how light varies with angle and

each choice is expensive,” he says. As Pozzi

spectrum, in addition to space and time,

establishes and refines his algorithms,

rewiring of the brain’s system. Because parts

and how these attributes of light change

he also will develop methods to teach

of the brain are now dead, there are neurons

after interacting with a scene.

infrastructure planning and analysis.

“You can imagine stroke as a sort of

that contribute completely differently to that

“With this award, my team will develop

“Managers have to compare the benefits

Partnering with CMU’s Summer Engineering

circuit,” says Chase, who also is a member

imaging systems and associated algorithms

Experience for Girls program, Pozzi plans to

of the Center for Neural Basis of Cognition,

for sensing and interpreting interactions of

build a simulation game in which students

a joint project between Carnegie Mellon

light with materials. The visual complexity

act as virtual infrastructure managers who

and the University of Pittsburgh devoted to

of real-world scenes implies that these

must develop, test, and revise decision-

investigating the neural mechanisms that

interactions are often intricate and complex,”

making strategies in the face of persistent

give rise to human cognitive abilities. “In

says Sankaranarayanan. “Our goal is to study

risk and uncertainty.

stroke rehabilitation, the brain must learn to

these interactions at unprecedented space

use those neurons in an appropriate way for

and time resolutions, thereby advancing

long-term project that allows me to

this altered system. We want to understand

research in many disciplines including

investigate topics I am passionate about, to

how the brain does this learning.”

computer vision, graphics as well as 3-D

educate students and to form a path in the

acquisition and printing.”

direction in which I want to research and

A major challenge in studying skill learning is that most movements engage tens of thousands of neurons, and the

One goal of the project is to develop

“I’m excited because it’s an expansive,

teach,” he says.

robust 3-D scanning of objects, which will ENGINEERING ENGINEERING SPRING FALL 2017

18 22

Carnegie’s Medallion In May 2017, alumnus Thomas Bugel (ChemE’55) sent the College of Engineering a remarkable treasure—a historic medallion he obtained while

INI Director Receives Professorship In Information Networking Dena Haritos Tsamitis, director of the Information Networking Institute, is the first recipient of the newly established Barbara Lazarus Professorship in Information Networking.

Carnegie Mellon created the professorship to celebrate the legacy of Barbara Lazarus, associate provost for academic affairs, teacher, scholar, and beloved member of the Carnegie Mellon community from 1985 to 2003. A nationally and internationally known scholar and activist, Lazarus worked tirelessly for the equity of women in the workplace, and the well-being and flourishing of graduate students and junior faculty at the university.

With over a decade of work fostering diversity at the INI, including

attending Carnegie Institute of Technology.

The medallion, engraved and signed by V.D. Brenner—you may know him as the sculptor who designed the U.S. Lincoln penny—commemorates the dedication of a new building to the Engineering Societies, courtesy of none other than our founder, Andrew Carnegie himself.

Bugel was unaware of the medallion’s significance when he found the artifact on campus, but for the next six decades, the medallion would act as a memento of his time at Carnegie Tech: “the best two years of college I ever spent,” says Bugel. Though much remains a mystery about the medallion—such as the means by which it arrived on campus, seeing as it was commissioned to commemorate the construction of the Engineering Societies’ Building and Engineers’ Club, erected in New York City in 1907—its significance to Bugel is clear. A reminder of his years at Carnegie Tech, Bugel thought of the beautifully carved 2”x3” copper rectangle as a good luck charm, carrying him throughout his long career in plastics, automotives, and environmental education.

Says Bugel: “It’s small, but it carries a lot of weight.”

the founding of Women@INI in 2005, Tsamitis exemplifies Lazarus’ commitment to promoting equality and inclusion.

“I believe in the importance of creating a nurturing and inspiring environment that promotes and celebrates gender and cultural equality, both within and beyond the INI,” said Tsamitis. “Barbara Lazarus was an early champion of the belief that culture matters. I am inspired by her example and honored to receive this professorship in her name.”

In fact, this is the second time Tsamitis has received a recognition honoring Lazarus’ legacy. In 2012, Tsamitis received the university’s Barbara Lazarus Award for Graduate Student and Junior Faculty Mentoring for her work with women and underrepresented minorities and her commitment to creating a culture that supports all.

“One of CMU’s greatest strengths is the caliber of our people,” said Dean James H. Garrett Jr., at a reception held to honor Tsamitis on April 20. “It is my pleasure to celebrate the achievements of our friend and colleague Dena as she receives this incredible honor named after another legendary CMU woman.”

1907

PPP: The Strongest Team in DefCon History At a time when the need and demand for

“More now than ever, the skills used in

cybersecurity expertise is at its highest,

this competition are becoming more relevant

CMU’s hacking team won its fourth “World

because cybersecurity is impacting all of our

Series of Hacking” title at the DefCon security

lives,” says Tim Becker, a team captain and

conference in Las Vegas, which took place

fourth-year computer science student. “It’s

on July 27 – 30, 2017. With four titles under

important that people have a place like this

their belt, the team has more wins than any

to hone their skills. The more we practice, the

other team in the 21-year history of this

better prepared we’ll be in the real world in

international competition.

dealing with actual cyberattacks.”

“The problem-solving skills required to

Becker launched his hacking career

win these contests mimic those needed

as a high school student in 2013 after

by governments and businesses alike to

participating in CyLab’s “picoCTF,” an online

anticipate and prevent cyberattacks,” says

capture the flag competition for middle and

David Brumley, the director of Carnegie

high school students.

Mellon’s CyLab Security and Privacy Institute, and a faculty advisor to the team.

This year’s competitor field consisted

“When I was a high school senior, I was competing with some friends and we didn’t expect to do very well. But after the first day

of 15 teams from more than 10 countries.

of the competition, we were in the top 10,”

To earn a spot at the table, each team had

Becker says. “That’s when I realized, maybe

to win a series of qualifying competitions

we’re pretty good at this.”

that were held over the past year. Carnegie

Becker’s team ended up finishing third

Mellon’s team, the Plaid Parliament of

overall in the 2013 competition, and that set

Pwning, was granted a spot for being last

him off to study computer security in college.

year’s champions.

Four DefCon wins later, Becker is on track

For the past 21 years, teams of hackers from scores of different countries around

towards a career in cybersecurity.

The Carnegie Mellon hacking team formed

the world have qualified for and competed

in 2009 and began competing in DefCon’s

in DefCon’s digital “Capture the Flag”

Capture the Flag competition in 2010. The

competition. Over the course of the three-

team previously won the contest in 2013,

day competition, teams try to break into

2014, and 2016.

competitors’ servers while protecting their own. During successful breaches, teams grab virtual “flags” and earn points.

ENGINEERING FALL 2017

24

Carnegie Mellon University has named Jay F. Whitacre, professor of MSE and EPP, the new director of the Wilton E. Scott Institute for Energy ENGINEERING.INSIDE THE COLLEGE

Innovation. On May 1, 2017, Whitacre succeeded Director Jared L. Cohon, CMU President Emeritus and University Professor of CEE and EPP. From the start of his career at the Jet Propulsion Laboratory, where he studied electrochemical energy storage and conversion, Whitacre has been a standout innovator in the world of energy technologies.

He came to CMU in 2007, where he quickly made a name for himself in the field of low-cost stationary energy storage by developing a novel battery that uses non-toxic, non-flammable chemicals. This unique battery became the basis for Aquion Energy, an energy storage technology company Whitacre founded in 2009.

As an institutional leader, Whitacre has served on the boards of multiple energy technology companies, as well as on committees for the National Academies of Sciences, Engineering, and Medicine. In 2015, he received the prestigious Lemelson-MIT Prize for inventing the first mass-produced low-cost, eco-friendly battery.

“In today’s energy-constrained world, it is more important than ever for Carnegie Mellon to remain on the cutting edge of research, education, and innovation in energy technology and policy,” says Provost Farnam Jahanian. “Bringing Jay’s wealth of experience in energy research and business development to the Scott Institute will ensure CMU continues to play a leadership role in this key strategic area.”

The Scott Institute was established in 2012, thanks to the generosity of CMU alumni Sherman Scott (E’66), president and founder of Delmar Systems, and his wife, university trustee Joyce Bowie Scott (A’65).

Their founding gift, along with additional support for energy research provided by the Richard King Mellon Foundation of Pittsburgh, is what has made the Institute’s groundbreaking work possible.

The Institute’s mission is to promote research that improves energy efficiency, creates innovations in energy technologies and policies, and broadens the world’s mix of energy sources in a way that is sustainable, reliable, and affordable.

Cohon, who had served as director since 2014, says, “I decided that this was a good time to pass the directorship of the Scott Institute to the next generation of emerging faculty leaders. We have created a firm foundation of programs and staff on which real progress and growth can be built, and I think Professor Whitacre is an outstanding choice to lead the Institute into its exciting future.” Cohon assumed the position from founding director M. Granger Morgan, whose focus was on connecting CMU research to public policy, and establishing the Institute during the building of Scott Hall.

INNOVATOR Leads Energy Institute

There are over 130 professors currently listed as affiliates of the Scott Institute, either by research, academics, or other energy-related activities at CMU.

Seeding Energy Research The Wilton E. Scott Institute for Energy Innovation announced the

• A team of faculty researchers led by Assistant Professor Venkat

2017 recipients of its annual Seed Grants for Energy Research,

Viswanathan (MechE) will develop a catalytic approach to improving

which supports efforts across the university in the areas of energy,

the long-term stability of lithium-ion batteries by building a bridge

environment, and policy.

between understanding electrocatalytic oxygen evolution and oxygen

Founded in 2013, this seed grant program has funded nearly 40

release in high voltage cathode materials. The aim is to remove

research teams. The 2017 funding alone totals nearly $553K from the

obstacles, such as cost and limited storage capacity, that limit electric

Scott Institute and the EQT Foundation. The EQT-funded projects will

vehicle adoption.

specifically research natural gas-related issues.

• Associate Professors Athanasios Karamalidis (CEE) and Newell The following teams from the College of Engineering were funded:

Washburn (Mellon College of Science, Chemistry) will conduct research to develop a continuously operating, small-scale, field-ready separation

• Associate Professor Shawn Litster (MechE, MSE), along with Associate

system that can be deployed at sites that produce waste water—such

Professor Kevin Noonan and Professor Tomasz Kowalewski (Mellon

as gas operations, geothermal utilities, coal power plant effluents, and

College of Science, Chemistry) aim to advance the science and

more—and extract rare earth elements from the water produced.

engineering of alkaline membrane fuel cells (AMFCs) by increasing the

These elements are necessary for the development of current and next-

conductivity and durability of the membranes, in hopes of positioning

generation green energy technologies. Funded by EQT Foundation.

AMFCs as an attractive alternative to common proton exchange membrane fuel cells.

• Professor Jay Whitacre (MSE, EPP) and Associate Professor Meagan Mauter (EPP, CEE) will create a Concurrent Assessment and Design

• Professor Jeremy Michalek (MechE, EPP) and co-PIs Professor Inês

of Systems (CADS) platform for assessing the complex intersections

Azevedo (EPP) and Assistant Professor Constantine Samaras (CEE) will

between policy, technology, and human responses to system

examine the impact of autonomous taxis and shared mobility services

perturbation in climate change mitigation and adaptation. The

such as Uber and Lyft on energy consumption, vehicle use, and

proposed work balances the need for rapid, quantitative, and intuitive

greenhouse gas and criteria air pollutant emissions.

models by hybridizing these existing models into the CADS platform.

Funded by EQT Foundation.

ENGINEERING FALL 2017

26

Honoring Angel Jordan

HONORING JORDAN’S LEGACY The Angel Jordan Early Career Professorship In spring 2017, the College of Engineering celebrated the establishment of the Angel Jordan Early Career Professorship. The fund was established through gifts made by more than 50 alumni and friends grateful to Dr. Jordan for his leadership and legacy. The Jordan Professorship is awarded to an outstanding faculty member in ECE. In spring 2017, it was awarded to Vyas Sekar. If you would like to honor Angel Jordan’s legacy at CMU, please visit Giving.cmu.edu/AngelJordan

At a time when American engineering

thought leaders in computer engineering,

universities were just taking shape, in the

engineering design, magnetics, and systems,

1970s, the Carnegie Institute of Technology

and they transformed Carnegie Mellon into

was forging ahead. With Angel Jordan at the

the engineering powerhouse that it is today.”

helm, the university became synonymous

Serving in various roles throughout

with excellence, innovation, and global

his career, Jordan left his mark on the

prominence. His 40+ year career directly

university. After earning his Ph.D., he

coincided with the remarkable rise of

became an assistant professor of electrical

Carnegie Mellon University.

and computer engineering in 1959, was

Angel Jordan, trailblazer, key architect,

made an associate professor in 1962, and

and builder of a department, a college, and a

became a full professor in 1966. After

university, passed away on August 4, 2017.

serving as dean of the Carnegie Institute

“Angel Jordan helped to define what

of Technology from 1979-83, Jordan was

Carnegie Mellon is today,” said James H.

the university provost until 1991. He had a

Garrett, Jr., dean of the College of Engineering.

pivotal role in launching many powerhouse

“He had vision and he had the ability to

programs and institutions, including the

bring amazing people to Carnegie Mellon

School of Computer Science, the Software

to achieve that vision. Those people were

Engineering Institute, and the Robotics

JORDAN’S EARLY CAREER PROFESSORSHIP Assistant Professor of ECE Vyas Sekar was awarded the Angel Jordan Early Career Professorship for his contributions in the academic and research sectors of Carnegie Mellon’s College of Engineering.

The Department of ECE created the Angel Jordan Early Career Professorship to honor and celebrate the undeniable impact that Angel Jordan has had on Carnegie Mellon and the College of Engineering.

Representatives from the Department of ECE created the Angel Jordan Early Career Professorship to bestow appreciation upon a young faculty member who has achieved great feats in the beginning of their career. By honoring a faculty member with this professorship, the ECE Department indicates the topnotch quality of its instructors. Assistant Professor Vyas Sekar was recognized Institute. Upon his retirement in 2000,

sure faculty felt comfortable. He treated us

Jordan remained a friend and confidant to

like family.”

many of Carnegie Mellon’s leaders.

“Angel was a constant presence on

As a researcher, Jordan made numerous scientific and technical contributions in

campus deep into his retirement,” said

semiconductor electronics and materials

Jelena Kovačević, Hamerschlag University

science and engineering, including tunnel

Professor and head of the electrical and

diodes, junction devices, photodiodes, and

computer engineering department. “He was

high frequency devices, to name a few. His

always willing to lend an ear and listen. I

work in these areas formed the foundation for

was particularly touched that he would take

many research areas in today’s Department of

the time to stop by and see how I was doing

Electrical and Computer Engineering.

as a new department head and offer advice.

I am heartbroken.”

A dedicated and inspirational department

Angel Jordan is an icon at Carnegie Mellon University. Dedicating his entire career to the university, he was one of the leaders

head, Jordan often went out of his way

who transformed Carnegie Mellon into the

to make his faculty comfortable. Upon

prominent world-class educational and

Professor Dan Siewiorek’s first day teaching

research institution it is today. His landmark

in the electrical and computer engineering

achievements will have a resounding impact

department, Jordan volunteered to co-teach

for years to come.

his first class.

“I showed up one wintry January day and

Born in Pamplona, Spain, Jordan attended the University of Zaragoza where

was scheduled to teach a Monday morning

he received his undergraduate degree

8:30 a.m. junior-level course on computer

in Physics in 1952. After coming to the

architecture. To give me confidence, Angel

United States, Jordan enrolled at what was

volunteered to be my co-instructor,” said

then the Carnegie Institute of Technology,

Siewiorek. “The area was outside of his main

where he earned a Master of Science and

field of expertise, but Angel wanted to make

Ph.D. in Electrical Engineering in 1959. His

sure my first day in the classroom went

wife, Nieves Jordan, received her Master

smoothly. This was just one of the many

of Science and Ph.D. degrees in Electrical

things he did throughout his career to make

Engineering in 1958 and 1960.

as one of these influential instructors because of his profound research on networking, security, and systems, as well as his undeniable teaching abilities.

During the award ceremony held on May 11, 2017, Sekar commented on the impact that Jordan had on his own career: “It is a great honor as a junior faculty member to receive a named professorship in general, especially so when it is named after one of the stalwarts of Carnegie Mellon. As I was reading through Angel’s illustrious career at CMU and beyond, it struck me that in many ways, I couldn’t even be here receiving this award if not for many of his contributions in the past.

He was instrumental in the creation of the School of Computer Science, which I graduated from, and many of the other institutions here that I benefited from in graduate school. It is indeed very humbling to look back at Angel’s amazing technical and service contributions to CMU and the broader engineering and computing community, and it gives me renewed motivation and drive to try and live up to being a worthy recipient of a professorship named in his honor.”

ENGINEERING FALL 2017

28

ENGINEERING.STUDENTS

Breast cancer is the most prevalent nonskin cancer among women. Starting as an abnormality within the breast, the cancer is at risk of spreading to the rest of the body; therefore, early detection of such abnormalities is paramount. Equally as important as detecting them is monitoring them regularly for changes—but do we have the tools necessary to do this?

The answer is yes—and no. Most of the procedures for diagnosis and monitoring— mammograms, breast ultrasounds, breast biopsies, breast MRIs—require expensive equipment. It’s not feasible for these procedures to be performed regularly, or in underdeveloped areas where access to such equipment is rare. Although breast self-examinations are a highly advocated, inexpensive way to detect lesions, it does not provide information that can be recorded and tracked.

Thus was born the need for an alternative method—and responding to that need is BME Ph.D. student Constance Robbins, who is working with Jana Kainerstorfer, assistant professor of BME, to develop a new device.

James Antaki, professor of BME, is also a collaborator on this project.

“We’re not trying to replace mammograms,” says Robbins. “If you have results that are inconclusive, but there’s not high enough of a risk to justify an immediate biopsy, then this could be something we could use this device for.”

The first iteration was called “Palpaid,” a project developed by Molly Blank, a former BME Ph.D. student advised by Antaki. Palpaid was intended to be a handheld device that measured the mechanical properties of a lesion. The device relied on the stiffness of the lesion compared to its surrounding tissue. When pressed against the breast, the device’s flexible, reflective lens would deform around the stiff tissue and capture a topographic image of the deformation, which could then be examined for information on size, shape, stiffness, and location.

A HANDHELD DEVICE TO MONITOR FOR BREAST CANCER

Robbins decided to take Palpaid

“If the lesion is too deep, we won’t be able

further and measure vascularization,

to detect it—that’s why we use compression

too. Breast lesions are stiff and highly

to decrease the distance from the lesion to

vascularized, meaning they contain greater

our sensors,” says Robbins.

total hemoglobin concentration than the

By compressing the tissue and essentially

surrounding softer tissue. But how can

pushing the lesion closer towards the surface

something like blood content be measured

of the breast, the light can penetrate further

non-invasively?

into the tissue, and the device can get a better read on the lesion’s blood content.

BME Ph.D. student Constance Robbins is creating a breast cancer lesionmonitoring device that has the potential to revolutionize breast cancer treatment.

The answer is with light. Robbins is

Robbins presented her research at

using an imaging method called spatial

the SPIE Photonics West Conference, a

frequency domain imaging (SFDI), where

conference dedicated to optics, in January

different patterns of light are shone on an

2017. She will finish her required Ph.D.

area to extract information from the way

coursework next year and will then devote

those patterns are reflected differently. The

her time entirely towards research.

variations in reflection tell us how much light

Having already started the next round of

the tissue can absorb, which in turn depends

experiments, Robbins hopes to create a

on the concentration of hemoglobin in the

consumer electronic device that has the

tissue. However, light can only penetrate so

potential to revolutionize breast cancer

far past the surface of the breast—what if a

treatment worldwide.

lesion is too deeply embedded?

LEFT: Model of imaging and compression system that Robbins demonstrated at SPIE conference. Here, a petri dish is placed underneath an acrylic plate. In the top left, a CCD camera is mounted and pointed at the petri dish. In the top right, a laser projector also points down.

RIGHT: Demo of technology used in consumer elctronic device

ENGINEERING FALL 2017

30

Self-Driving Buggies Make History For the first time in history, two self-driving buggies took part in Carnegie Mellon’s Spring Carnival Sweepstakes. CMU has been at the forefront of autonomous driving technology for three decades, and students have been working to bring that technology to the buggy races. This year they succeeded.

In April, during the early morning practices, two buggies driven by computers, rather than by students tucked inside the torpedolike shell, successfully steered and braked their way around the 4,412-foot-long course.

On April 8, team RoboBuggy fielded the first autonomous buggy to steer itself through the course. And on April 9, the Atlas Project, part of the Carnegie Involvement Association, also known as CIA, completed its own first successful autonomous run.

Like traditional buggies, the machines are pushed uphill by humans on Tech Street, then freeroll down Schenley Drive before being pushed up Frew Street. The difference is that the buggy itself is doing the steering and navigation throughout.

Both self-driving buggies were displayed during the Sweepstakes Design Showcase and Buggy Races as part of Spring Carnival 2017.

“I joined RoboClub, which is the organization that sponsors RoboBuggy, in my freshman year,” said Danielle Quan, chairman of RoboBuggy and a junior majoring in MechE and robotics. “This is very quintessentially CMU—it’s buggies and robots, so I immediately wanted to get involved.”

Matt White, who graduated from Carnegie Mellon’s School of Computer Science in 1996, began the RoboBuggy project during a conversation back then with Mark Stehlik, now associate dean for outreach in the School of Computer Science. They had the idea of adapting the technology coming out of CMU’s Field Robotics Center’s NavLab project for Buggy. He worked on a hardware platform as an independent study.

Since then, RoboBuggy has been revived by several interdisciplinary student teams. The most recent iteration of RoboBuggy has existed since 2013 as a project within the CMU Robotics Club. The Atlas Project, a separate autonomous buggy team, started two years ago.

“It was surreal, honestly, watching it steer itself after having failed every other time,” said Benjamin Warwick, founder of the Atlas Project and a junior in MechE. “I’m hoping this will encourage other people to start robotic buggies of their own.”

The first buggy races were held as part of the first alumni celebration, called Campus Week, in 1920. In 1928, Campus Week was replaced by Spring Carnival and a booth competition was introduced. Frew Street was completed and the course took on its present-day format.

INI Students Innovate At Internships From Pittsburgh to Palo Alto and Fort Meade to Seattle, Information Networking Institute (INI) students innovated at summer internships with top tech companies, research centers and government agencies. While the perks are many—dog-friendly offices, outings to baseball games, and lots of free food—the key takeaway was working independently on a project with real impact in the fields of data science, mobility, machine learning, software engineering, cybersecurity, R&D, and more.

At Facebook’s 1 Hacker Way, Kung-Hsien Yu was told to “move fast and don’t be afraid to break something.” It is a sentiment that resonates across all internships as INI students applied classroom knowledge to real-world projects. Amidst rapid prototyping, building software, pushing code into production, and developing security tools, students gained invaluable work experience and networking connections that laid the groundwork for full-time employment.

As a software engineering intern at Pinterest, Caleb de la Cruz endeavored to make the virtual pinboard faster. “I love this job because you get to work on challenging problems alongside coworkers who are extremely talented,” he says.

In the shadow of San Francisco’s Bay Bridge, research intern Simran Gujral modernized the TLS stack in Mozilla’s Servo. Along the way, she learned that failure and asking questions are both part of the creative process. “Each person here is a creator,” she says. “They have amazing ideas and the design space is so open.

We are constantly pushed to innovate.”

At LinkedIn, Zhiyuan Lin attended the Hadoop Summit, where she networked with the industry’s best engineers and managers. Her project—automating the Apache Spark tuning recommendation—will be an incredible resource for LinkedIn and the entire Spark community.

This year’s top internship employers were Amazon, VMware, Yahoo, Google and Salesforce. Several students remained in Pittsburgh, interning with CyLab and the Software Engineering Institute’s CERT division.

“I am so proud of our students for stepping outside of their comfort zones to connect with top employers,” says Jessica Budik, INI’s associate director of career services. “By obtaining these outstanding career opportunities and developing new skills, our students make a difference in the tech world.”

ENGINEERING FALL 2017

32

ENGINEERING.ALUMNI

Father Like Son In 2010, Kelly was unsure how to pursue

integrated into subassemblies where precise

his interest in education. At the age of 58,

measurements are taken to ensure they are

he joined Teach for America. However, while

meeting the demanding requirements of

teaching middle school science in Southern

REO’s customers. The company was recently

Mississippi, Kelly realized that his talents

visited by U.S. Secretary of Labor Alexander

would be better applied elsewhere. Instead

Acosta as part of the administration’s interest

of teaching in a classroom, he decided

in growing the apprenticeship initiative

to expose high school students to career

across the country.

options in his line of work—optics.

“You don’t have to go to Southern

Integrating itself into the community, REO judges high school science fairs and sends

Mississippi to do something great for

female engineers to schools during Women in

underserved kids,” says Kelly. “You can do it

Engineering week. High school teachers visit

anywhere if you run a company well, do the

the company to get a better understanding of

right things for the culture, and mentor kids.”

what high-tech manufacturing jobs entail. REO

Kelly returned to Boulder as CEO of REO, a

also participates in Backpacks to Briefcases, a

precision optics manufacturing company. REO

yearly luncheon that focuses on the transition

manufactures optics for use in industrial and

from being in high school to going into

medical applications, such as metal cutting

business.

and drilling, tattoo removal, and Lasik surgery.

Spreading his commitment to STEM

In addition, the company provides optics that

education beyond Boulder, Kelly established

are used in defense applications: 35 optics in

a $50,000 fund, the Kenneth J. and Paul C.

the F35, and a number of applications for laser

Kelly Technology Outreach Fund, at Carnegie

targeting and electronic countermeasures.

Mellon. This fund supports several projects,

Kelly worked with Front Range Community College in Boulder to design a special program for recent high school graduates.

including a large STEM outreach event that occurred last winter.

The fund is named in part for Kelly’s late

They geared the program toward minority

father Kenneth, who ran the materials lab

students, those who would be the first in

at Pratt & Whitney Aircraft, and worked on

Most companies don’t work with local

their family to graduate from college, and

the Manhattan Project in WWII at the Oak

colleges to educate minority students or

others who were not positioned financially or

Ridge National Labs. The influence the elder

invite students to their facilities for hands-

academically to go to college.

Kelly had on his son is evident. Kenneth,

on training. Most companies are not run by

Students in the program spend half of their

who earned a master’s degree in chemistry,

Paul Kelly, College of Engineering alumnus

time in class and the other half getting hands-

stirred Paul’s interest in metallurgical

and CEO of REO, Research Electro-Optics, in

on training in REO’s state-of-the-art facility.

engineering. When Kenneth retired, he

Boulder, Colorado.

REO’s employees, 13 of whom have Ph.D.s,

tutored kids in math and science. Continuing

carry the teaching load for these classes. If the

the tradition, Paul’s son, Brian Kelly, is a

STEM (Science, Technology, Engineering,

student decides to go beyond the Associate’s

middle school teacher in an economically

and Mathematics) education because he

Degree, REO reimburses 100% of their tuition

deprived neighborhood in South Chicago.

thinks that kids are not aware enough of the

costs, as they do for all of their employees.

Kelly has always had a deep interest in

possibilities that exist in these fields.

“I believe very strongly that each of us

“I wanted to honor my father by setting

REO has invested in six apprentices to

up this fund that would carry on his mission,

date. When these apprentices are not taking

by exposing and supporting kids who have an

has a skill, a capability, a talent inside of us,

optics classes at the community college,

interest in technology and science careers,”

and it’s just a matter of finding that,” says

they are working in the manufacturing

says Kelly. “It’s our responsibility—those of us

Kelly, who has a master’s degree from CMU in

process in which optics are fabricated, a thin

in this profession—to expose kids to the fun

metallurgical engineering.

film coating is applied, and the optics are

and beauty of science.”

DAN STREYLE

Give strategically, Support generously.

DAN STREYLE (CEE'75) sets his focus

Learn how easy it is to achieve

on the future. As lead project engineer,

your philanthropic vision through a

he facilitated the building of the

planned gift. Contact the Office of

University of Phoenix Stadium, where

Gift Planning today at 412.268.5346

the Arizona Cardinals play, creating

or askjoebull@andrew.cmu.edu, or

an impressive structure that will stand

visit giftplanning.cmu.edu.

the test of time. At Carnegie Mellon University, Dan has created a lasting legacy through a gift in his estate plan that will benefit undergraduate engineering research, spurring innovation for generations to come.

College of Engineering

College of Engineering

Office of the Dean College of Engineering Carnegie Mellon University Pittsburgh, Pennsylvania 15213-3890

Nonprofit Org. U.S. Postage PAID Permit No. 251 Carnegie Mellon University does not discriminate in admission, employment, or administration of its programs or activities on the basis of race, color, national origin, sex, handicap or disability, age, sexual orientation, gender identity, religion, creed, ancestry, belief, veteran status or genetic information. Furthermore, Carnegie Mellon University does not discriminate and is required not to discriminate in violation of federal, state, or local laws or executive orders.

Inquiries concerning the application of and compliance with this statement should be directed to the vice president for campus affairs, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, telephone 412-268-2056. Carnegie Mellon University publishes an annual campus security and fire safety report describing the university’s security, alcohol and drug sexual assault, and fire safety policies and containing statistics about the number and type of crimes committed on the campus and the number and cause of fires in campus residence facilities during the preceding three years. You can obtain a copy by contacting the Carnegie Mellon Police Department at 412-268-2323. The annual security and fire safety report is also available online at http://www.cmu.edu/police/annualreports.