Reaping the Benefits of a Diverse Background

Dina Katabi, recipient of the 2017 ACM Prize in Computing, took a winding road to computing, and it paid off. Now a professor of electrical engineering and computer science at the Massachusetts Institute of Technology (MIT), Katabi began her career in medicine. Since making the transition, she has made numerous creative contributions to wireless network design. Today, she is helping to develop medical applications for a technology she pioneered, which uses wireless signals to sense humans and their movements through walls—her early training coming full circle.

Your undergraduate degree is in electrical engineering, but you began by studying medicine.

In Syria, after high school, there is a nationwide exam, and the expectation is that the top people will go to medical school. I took the exam, and I ranked very high. I also come from a family of doctors. So I went to med school, but after the first year, I decided I could not continue. I wanted to do math and engineering, so I decided to switch.

You then came to the U.S., and did your Ph.D. in computer science.

At the time, computer science (CS) was a very new field in Syria. In fact, at the school I attended, there was no such thing as a CS school or department. But I was always fascinated with computers, and when I came to the U.S., I wanted to learn more about algorithms.

Having experience with different fields seems to have proven beneficial to your work.

I’ve benefited from having a very diverse background, which has enabled me to see beyond the field I am in. Particularly when I was working on wireless systems, my background gave me the expertise I needed to design the circuit, the signal, and also the algorithm that extracts information from that signal. You can design many systems with electrical engineering, but the ability to add intelligence to them using CS is much more powerful than if it was just pure signal processing.

In some of your earliest work at MIT, you collaborated with David Clark—the Internet’s chief protocol architect during most of the 1980s—on network control, where one of the biggest problems is managing transmissions when they threaten to overwhelm the network.

At the time, the traditional method of congestion control was based on heuristics. It was more of an art than anything. But it was often not very efficient—not very fair to different users—because the Internet is just too big. With my thesis, I tried to connect that art and intuition to the field of control theory, which is a subfield in electrical engineering that is typically used to control plants and manufacturing systems. So you can keep the intuition, but if you infuse into it some of the mathematical models, you can achieve much better results. You can make the network more stable and achieve more efficient systems.

After you received your Ph.D., you stayed at MIT and began working with information theorist Muriel Médard on network coding, a technique for increasing networks’ data capacity that was promising in theory, but had not yet been shown to work on a real network.

When we began our work, network coding had shown high gains in specific examples, but those examples did not map to the way that networks really operate. For instance, the theory of network coding was defined in the context of something called multicast.

Multicast is a communications protocol in which you deliver the same information to a group of destinations simultaneously.

But in networking, typically, that’s not how it works. In networking, you typically have unicast, where one sender transmits to a single destination. Even when you are sending something like broadcast television over the Internet, your broadcast is actually using unicast. You have your server turning that traffic to all the individuals who are interested in it.

What Muriel and I did was try to take that really beautiful, elegant theory, and think about it in the context of real networks. I felt wireless networks, in particular, might be the right environment for this technology. Wireless is way more limited in terms of data rate and bandwidth than wired networks, and it’s also less reliable. So network coding is an ideal solution when you make an error in your transmission.

In your recent work, you’ve used wireless signals to track people’s motions—even through walls. How did you get that idea?

When we began, it was really curiosity. Let’s say there is a room and you don’t have access to it. Can you tell if there are people in the room? If you can tell there are people, can you tell how many people? When we tried that, we didn’t really know whether or not it was possible, and we certainly didn’t know what kind of application you’d use it for. All we knew is that we have been able to track people using their cellphones—so, using a wireless signal, but a wireless signal that is emitted from a device. And we have some understanding of how wireless works in an indoor environment and propagates through walls and materials.

After your initial demonstrations were successful, the questions got more complex, and practical applications began to present themselves.

Once we started working with it, we began to have all these ideas—why stop at just being able to see if people are moving? Can you tell how they are moving? Can you tell how many people there are? It turns out you can, because they are breathing… So what other physiological signals can you extract?

And these questions are extremely intellectually interesting, but it’s not just that; they have very practical and useful applications to people’s lives!

You’re now working, through a start-up called Emerald, to commercialize the technology and develop some of those applications—for instance, remotely monitoring people’s health.

We talk a lot about the smart home, but really the smartest thing a home can do is to take care of us and our health. Our vision is to have a technology that disappears into the environment; I don’t have to enter information about my heartrate, or put some device on myself and remember to charge it. I don’t need to change my behavior in anyway, but still there is a home that’s watching over my health and keeping track of problems early on—or even before they occur—and alerting doctors or the hospital or a caregiver.

That sounds promising. Where are you in your efforts?

At this early stage, our focus is to work with healthcare providers, on the one hand, and with the biotech and pharma industry, on the other. It turns out there are many deep physiological signals we can extract, so we need to connect with people who understand what those signals mean in the context of diseases. I can tell you that my mom is walking well or that she fell—that’s the extent of it. I couldn’t tell you if the patterns of information indicate we should change the dose on her Parkinson’s medication.

One of the most consistently cited features of your work is creativity.

In general, in almost all the stuff I do, I’m driven by curiosity. I’m always interested in trying something where I don’t know the answer, or where I’m not sure whether the answer is "yes" or "no."

Back to Top