Jayant Kalagnanam

Master of time. Optimization researcher. Modeling guru.

March 23, 2021

Marshalling science and technology to build efficiencies into heavy manufacturing, infrastructure projects, and cities, Dr. Jayant Kalagnanam’s work deploys AI and ‘the science of better’ to make the world run more smoothly. Kalagnanam was selected to be part of the Industry Academy’s inaugural class of Distinguished Industry Leaders, prestigious thinkers who are globally recognized for their industry-transforming work and leadership.

A lot of the work you do has an outsized impact on our everyday lives, but it's invisible to most of the world — tiny sensors embedded in heavy equipment and infrastructure. How does it feel to have such a behind-the-scenes job?

I never thought of it that way, but typically in research we are behind the scenes. We provide the AI techniques and platforms for different uses, and we build out a lot of first-of-their-kind projects. iIn that context I do get a fair amount of opportunity to be with clients. It’s not all behind the scenes.

Did you ever imagine you'd be doing this sort of work? What did you want to be when you grew up?

I always wanted to be an engineer, even if my parents would have liked me to have gone into medicine. They weren't too happy about that. But I always had a strong leaning toward engineering and modeling. To a large extent, right through my studies I gravitated toward modeling various physical systems. The best research problems are defined when you're trying to solve real-life problems, things like figuring out how to incorporate more renewable energy into our current electrical grids, or maximizing efficiency in the steel industry to minimize waste and energy use. A lot of these problems have very real applications.

Your work essentially helps humans avoid hiccups in the regular flow of business, yes?

Typically, most of my work is to improve efficiency. In the optimization research community we often brand ourselves as ‘the science of better.’ Everything we do, the application of optimization research—including with AI—is all about targeting inefficiencies and using technology to make the world a better place. If you can improve the efficiency of processing aluminum by even a percent, the energy you save can be massive. That reduces a carbon footprint and improves the cost of things you're manufacturing.

What's the most fun part about working to make cities smarter with technology?

The world population is getting more and more concentrated into cities. If the population of a city doubles, you're not going to be able to just build a road to deal with it. It becomes a puzzle of how to use and design infrastructure more efficiently. Essentially, the fun part is giving more information to people in cities to make decisions on how to navigate what they do on a day-to-day basis.

I was in Singapore for a couple years for IBM and a lot of the work we initiated was around improving roadways. We also did a lot of work around environmental issues, using satellite and other sources of data to measure air quality in a city and alert people to upcoming issues. In Singapore smog was drifting in from other parts of Asia, for instance.

Another fun project is something we’re doing in the IBM labs in Zurich, Switzerland and Haifa, Israel, where we’re training drones to go inspect things like bridges. They can strike out on their own, collect imagery of cracks on bridges, send those images back and be analyzed to figure out if work needs to be done. Not having to send a person out to physical sites saves a lot of time, energy, and money.

What's the hardest part?

All these applications tend to be an exercise where you have to take AI technology where it is today and make it work to solve a problem of the future.

Think about the drones that take photos of bridges. When we think of drones, we might think of someone with a remote in their hands. But the challenge is to make this process more autonomous, so a drone can just take a flight path and go where it needs to go and take the pictures it needs to take. That goes beyond what typical AI would do, and so there are challenges there. And when images do come back, they’re shots of cracks in a cement or concrete structure, not classic images we’ve typically used for classifying and analyzing. We have to actually be able to identify these cracks, measure them, and then know to send out an alert. In order to do all this, we really need to stretch AI into a new place, and that requires new techniques and training deep thinking models.

The other big body of work I've been doing for the last few years has involved working with oil and gas companies. I’ve been using the data that's continuously coming out of the manufacturing process and learning the behavior of these processes. We can then use those machine learning models to predict behavior and use optimization to figure out how we can improve efficiencies.

You've been published nearly 100 times, and you have your name on almost 50 patents. Now you've been named as one of IBM's Distinguished Industry Leaders. How did that feel?

Certainly, I felt honored and a little surprised to be in the inaugural class. It was humbling because IBM is a big company with an outsized societal impact with a lot of very talented people. One place you can really see this is at our conferences. I remember going to the THINK conference in 2019. I believe 35,000 people were at that event and so many of them were impressive. To be one of the first six chosen as a Distinguished Industry Leader was an honor. We’re all doing a lot of work on a day-by-day basis and we don't really sit down and think about that. So when someone recognizes you for it, it feels rewarding.