MIT Associate Professor Rafael Gómez Bombarelli has been using artificial intelligence to create new materials for more than a decade. As technology expanded, so did his ambitions.
Now, a newly tenured professor of materials science and engineering believes AI is poised to transform science in ways never before possible. His work at MIT and beyond is dedicated to accelerating that future.
“We are at the second tipping point.” Gomez-Bombarelli says. “The first one was around 2015, when we saw the first wave of representation learning, generative AI, and high-throughput data in some areas of science. These were some of the techniques that I first brought to my lab at MIT. Now we’re in the second wave. I think we’re at an inflection point where we’re blending languages and integrating multiple modalities into general scientific intelligence. We’ll have all the model classes and scaling laws we need to reason about language, reason about material structure, and reason about synthetic recipes.”
Gomez-Bombarelli’s research combines physics-based simulation with approaches such as machine learning and generative AI to discover new materials with potential for real-world applications. His research has led to new materials for batteries, catalysts, plastics, and organic light-emitting diodes (OLEDs). He has also co-founded several companies and serves on the scientific advisory boards of startups applying AI to areas such as drug discovery and robotics. His latest company, Lila Sciences, is working to build a scientific superintelligence platform for the life sciences, chemistry, and materials science industries.
All of this work is designed to make future scientific research more seamless and productive than current research.
“AI for science is one of the most exciting and ambitious applications of AI,” says Gomez-Bombarelli. “Other AI applications come with more drawbacks and ambiguities. AI for science aims to bring about a better future over time.”
From experiments to simulations
Gómez-Bombarelli grew up in Spain and was drawn to the physical sciences from an early age. In 2001, he won the Chemistry Olympiad competition and studied chemistry as an undergraduate at the University of Salamanca in his hometown. Gómez-Bombarelli stayed to complete her PhD, studying the function of chemicals that damage DNA.
“My PhD started off experimentally, but somewhere along the way I got bit by the simulation and computer science bug,” he says. “I started simulating the same chemical reactions that I was measuring in the lab. I like the way programming organizes the brain. It felt like a natural way to organize thoughts. Programming is also much less limited by what you can do with your hands or scientific equipment.”
Gomez-Bombarelli then went to Scotland for a postdoctoral position, where she studied quantum effects in biology. Through that work, he connected with Alan Aspuru Guzik, a professor of chemistry at Harvard University, who joined him in 2014 as his next postdoctoral fellow.
“I was one of the first people to use generative AI in chemistry in 2016, and in 2015 I was part of the first team to use neural networks to understand molecules,” Gomez-Bombarelli says. “It was in the early, early days of deep learning in science.”
Gomez-Bombarelli also began an effort to eliminate the manual part of molecular simulations in order to perform more high-throughput experiments. He and his collaborators ultimately performed hundreds of thousands of calculations across the materials and discovered hundreds of promising materials suitable for testing.
After two years in the lab, Gómez-Bombarelli and Aspuru-Guzik founded a general-purpose materials calculations company that eventually focused on manufacturing organic light-emitting diodes. Gomez-Bombarelli joined the company full-time, which he says was the most difficult job of his career.
“It was great to create something tangible,” he says. “And I didn’t want to be a professor because I saw Aspul-Guzik running a lab. My father was a linguistics professor, so I thought it would be an easy job. Then I saw Aspul-Guzik leading a group of 40 people and traveling 120 days a year. It was insane. I didn’t think I had that kind of energy and creativity in me.”
In 2018, Aspuru-Guzik suggested that Gomez-Bombarelli apply for a new position in MIT’s Department of Materials Science and Engineering. However, Gomez-Bombarelli missed the deadline because she was worried about her teaching job. Aspr Guzik confronted him in his office, slammed his hand on the table and told him: “You need to apply for this.” That was enough to get Gomez-Bombarelli to put together a formal application.
Fortunately, Gomez-Bombarelli spent a lot of time at startup thinking about how to create value from discovering computational materials. During the interview process, he says he was drawn to MIT’s energy and collaborative spirit. He also began to realize the potential of research.
“Everything I did as a postdoc and at the company will be part of what I can do at MIT,” he says. “I was building products, and I still am. Suddenly, my world of work became part of a new world that I could explore and implement.”
It’s been nine years since Gomez Bombarelli joined MIT. His lab currently focuses on how atomic composition, structure, and reactivity affect material performance. He also helped develop tools that use high-throughput simulation to create new materials and blend deep learning and physically-based modeling.
“Physically-based simulations create data, and the more data you give them, the better the AI algorithms become,” says Gomez Bombarelli. “There’s all sorts of virtuous circles between AI and simulation.”
The research group he founded is solely for computational purposes and does not perform physical experiments.
“We’re very broad and can do a lot of things at once, so that’s a blessing,” he says. “We love collaborating with experimenters and trying to be a good partner to them. We also love creating computational tools that help experimenters prioritize ideas from AI.”
Gomez-Bombarelli also remains focused on real-world applications of the materials he has invented. His lab works closely with companies and organizations such as MIT’s Industry Liaison Program to understand the critical needs of the private sector and the practical hurdles to commercial development.
Accelerating science
As excitement around artificial intelligence explodes, Gomez-Bombarelli has seen the field mature. Companies like Meta, Microsoft, and Google’s DeepMind now regularly run physics-based simulations reminiscent of what he was working on in 2016. In November, the U.S. Department of Energy launched the Genesis mission to use AI to accelerate scientific discovery, national security, and energy dominance.
“AI for simulation has moved from something that could probably work to a consensus scientific opinion,” Gomez-Bombarelli says. “We’re at a tipping point. Humans think in natural language, we write papers in natural language, and we’re finding that these large language models that have mastered natural language have opened up the ability to accelerate science. We’ve seen that scaling works for simulation. We’ve seen that scaling works for language. Now we’ll look at how scaling works for science.”
When he first came to MIT, Gomez Bombarelli said he was shocked by how non-existent competition among researchers was. He is trying to bring the same positive-sum thinking to his research group of about 25 graduate students and postdocs.
“We have naturally grown into a really diverse group with diverse mentality,” says Gomez Bombarelli. “Everyone has their own career aspirations and strengths and weaknesses. I enjoy thinking about ways to help people become the best versions of themselves. Now I’m in a position to advocate for people to apply for faculty positions after the deadline. I feel like I’ve passed that baton.”
