Pooyan Jamshidi

Breakthrough Star: Pooyan Jamshidi

Lifelong interest in robotics leads professor to research in artificial intelligence, machine learning

Pooyan Jamshidi first became interested in robotics when he was about 10 years old, his imagination captured by the single book about the topic available in the small library in his elementary school in Iran. His childhood also included visiting his father at work, where he saw robots performing train manufacturing tasks on the factory floor.

That initial exposure was followed by picking up some magazines at a newsstand, his interest piqued by computer programming. He began to teach himself computer languages and coding. 

Those experiences started a career path for Jamshidi, who went on to earn his Ph.D. in computing at Dublin City University in Ireland, and after graduation, he started his first postdoctoral position at Imperial College London, and he eventually came to the U.S. to do his second postdoctoral position at Carnegie Mellon University.

He now is an assistant professor of computer science and engineering at UofSC and a visiting researcher at Google. Jamshidi’s research involves artificial intelligence, machine learning, and computer systems, developing novel algorithmic and theoretically principled methods to build reliable, high-performance machine learning systems, at Google scales. He is interested in applying the AI/ML algorithms in high-impact applications, including robotics, computer systems, healthcare, neuroscience, space explorations, engineering, and sciences.

The goal of his work is to develop the next generation of autonomous systems that can work in complex, real-world environments.

I have always been interested to see the impact of research going beyond the research lab and see the impact in the real world.

Pooyan Jamshidi 

Some of that “real-world” application is happening right now, as Jamshidi leads a collaborative team from students, researchers, and faculties from UofSC, CMU, UArk, and University of York (UK) that was awarded a $250,000 grant from NASA to use AI to increase the autonomy of future space missions. The two-year NASA Resource Adaptive Software Purpose-Built for Extraordinary Robotic Research Yields – Science Instruments grant (known as a RASPBERRY SI: https://nasa-raspberry-si.github.io/raspberry-si/) will develop novel machine learning systems to enable spacecraft to adapt and respond during deep space missions, something that has previously required human supervision and longer reaction times.

“We got lucky that our proposal was selected for funding by NASA as one of the two proposals out of the 17 submissions and got the chance to work on this exciting project to make the space lander autonomous by developing novel algorithms, novel methods, and approaches for machine learning to make it more intelligent,” he says. “There are a lot of opportunities, there's a lot of unknown unknowns there that even scientists don't know about them. So, if you send a spacecraft there remotely, and if something goes wrong, the mission could fail easily. And these missions are expensive. We want these space landers to be able to deal with unknown unknowns and many sources of uncertainties in planets like Europa moon, which are far away from the Earth, autonomously in a more intelligent way.”

“I have always been interested to see the impact of research going beyond the research lab and see the impact in the real world,” he says. “We closely work and collaborate with several researchers and scientists at NASA to evaluate our technology on physical testbed at NASA JPL and simulated environment at NASA Ames”  “This is one of the most exciting projects that I have ever been involved in.”

Jamshidi’s Lab, Artificial Intelligence and Systems Laboratory (AISys), in collaborations with collaborators, Christian Kaestner, associate professor at Carnegie Mellon University and Baishakhi Ray, Associate Professor at Columbia University are pursuing  alternative methods which are rooted in Causal AI for making a paradigm shift for testing and debugging complex machine learning systems with a $1.2 million grant from the National Science Foundation.

“While the importance of AI has become increasingly clear in recent times, understanding of how AI techniques can be utilized within complete systems that are reliable, secure, and robust has lagged behind. Jamshidi’s crucial research works to fill that gap,” says Jason O’Kane, professor and associate chair in the department of computer science and engineer. “His work is characterized by a unique blend of results that truly push forward the boundaries of understanding — in contrast to mere incremental improvements for specific cases — combined with rigorous and insightful analysis of those results.”

“I feel so lucky that I got the opportunity to work with so many amazing people around the world along with my research career. I wish we had more diversity in computer science; it is unbelievable that so many talented students, including female students, think that computer science and math are not for them; I can prove them wrong! the top students and researchers whom I got the chance to work with are all girls; I have learned much more from my peers, who are either female or come from a different culture than mine, I strongly believe diversity matters, and I wish one day come that we see orders of magnitude more students from underrepresented groups such as girls, blacks, Hispanics, Native Americans, and other underrepresented and under-served in the STEM enterprise.”