Professor Samuel Madden’s research aims to help anyone access different types of data through easy-to-use interfaces that let them query that data. “In general, I’m very driven by practical, real-world problems that I observe people having,” says Prof. Madden.
Prof. Madden, a Professor of Electrical Engineering and Computer Science in MIT CSAIL, directs initiatives in the space of Big Data and the applications of software systems to machine learning as well as machine learning to software systems. He also leads the Systems That Learn initiative and a research initiative on data science (referred to as “DSAIL”). “In both of these projects, we’re really focused on understanding how we can use technology, especially machine-learning technology, to prove software systems,” he says. Through these research initiatives, he is able to interact directly with a number of cutting-edge technology companies.
“One of the really great things about working with these companies is that you get exposed to their application domains and their problems, and they often introduce new problems to you that you didn’t know about before, that lead to advances in the underlying technology itself,” says Prof. Madden. “So, there’s kind of a two-way street where we can help the companies by figuring out how to apply the technology, and they can help us come up with new research ideas by bringing new problem domains to us that we might not have been familiar with before.”
As an example of an industry project that has come out of such a collaboration with CSAIL, Prof. Madden’s work on mobile sensor computing led to a project called CarTel, which stands for Car Telecommunications. “We took this technology and commercialized it as a company called Cambridge Mobile Telematics,” says Prof. Madden. “The vision of this company is to help make roads safer by making drivers better. Today, this technology is in use by several million drivers on the road every day. And people are changing their behavior, and that saves lives and makes roads safer.”
“I think in general that’s the type of work I tend to do, these very applied problems that are sort of driven by an observation of a real-world need,” Prof. Madden says. These observations lead him to a variety of new projects, such as a system for generating digital maps from satellite imagery and GPS traces.
“We observed a problem when we were in Qatar and we said, ‘I bet we can apply technology to solve this.’ We saw, for example, that if you tried to use Uber to take a trip in Qatar, you would often find that the driver couldn’t actually get to your destination because the quality of the maps were not updated frequently enough. So what we’ve been able to do in this project is to develop these new tools that take in satellite imagery and these GPS traces and convert them into up-to-date maps in a very timely fashion,” he says.
However, there are challenges involved in making digital maps, according to Prof. Madden. “When you’re looking at something like a satellite image, it’s hard to see the details of how the roads interconnect.” At a complicated freeway interchange, for example, “it’s very hard to tell from just a few pixels of an image whether those roads really intersect or not.”
That’s where the GPS traces come in. “Because the GPS traces the exact path that a particular car took, you can tell which roads are connected and which roads are not connected,” Prof. Madden explains. “What we’ve observed in this project is that we can use the satellite imagery to get the connectivity, the roadmap for the small roads, and then we can use the GPS traces to infer the connectivity of the more major roads that tend to be these more complicated interchanges that are hard to pick out from the satellite imagery.” He adds that he is excited about the possibility of democratizing maps in this project and helping people who are underserved by current mapping services.
This is just one of many projects in the works for Prof. Madden, who says that working in CSAIL has allowed him to address these types of real-world research problems head-on. “I get to work on different problems every day and I feel like I’m always learning and expanding what I know,” he says.
For companies that are interested in becoming industrial partners, he adds that CSAIL initiatives help in a number of ways. “I think the main reason is really to get access to the cutting-edge technology that we’re developing, to interact with the outstanding researchers and students, and be exposed to these new ideas that are coming down the pipe. There’s a ton of technology that’s happening at MIT, but it’s very hard for companies who are only observing the stream of research papers that are coming out to figure out how they might apply that technology to their own problem domains. So, I think many of our companies really benefit from being able to sit down with the researchers and students, talk about their problems, and then get this specialized advice about how to apply the technologies that researchers are developing into their world or their domain.”