Recent alums advance AI for science at startups

Using their computational training to apply artificial intelligence (AI) for science, chemical engineering alums see endless possibilities. Their work involves bringing together diverse data from experiments, simulations, and literature to create predictive models that can guide new experiments.

"Traditional experimentation and computation are often slow and expensive. By combining AI, automation, and simulation, we can explore chemical space far more efficiently," says Brook Wander ('24).

Chemical engineers are highly recruited in the field of AI for science because they are trained to think systematically about multi-scale problems. "Chemical engineering gives you macroscopic access to industry requirements. We study large-scale processes, economics of processes, system design, and fundamental sciences: math, physics, chemistry, sometimes even biology," says Adeesh Kolluru ('24).

Wander, Kolluru, Xiaoxiao (Lory) Wang, and Yuri Sanspeur are among the recent Ph.D. graduates working at the intersection of AI, computational chemistry, and materials science.

Xiaoxiao (Lory) Wang

Xiaoxiao (Lory) Wang ('25) is a research scientist at Lila Sciences. "The company mission is to empower researchers by providing advanced AI-driven tools and expert systems that streamline scientific inquiry, enhance experimental design, and accelerate the pace of discovery," she says. By combining generative AI with automated laboratories, Lila Sciences intends to design, run, and learn from experiments at scale. Wang is developing machine learning workflows for materials discovery.

The company is a good match for her passion for applying AI to real-world problems in science and also aligns with technical skills she developed during her Ph.D. Working with Zachary Ulissi and Rachel Kurchin, Wang used machine learning to accelerate atomistic simulations. She built specialized datasets, as well as an active learning workflow to guide experiments. At Lila Sciences, she applies similar approaches to integrate experimental data and computational simulation into AI workflows.

"Internships helped me understand the practical considerations that go into deploying machine learning models outside of academia," Wang says. During her Ph.D., she interned at Intel Labs and NXP Semiconductors, where she worked with large-scale datasets and real-world constraints. Hands-on experience with industry workflows, project timelines, and collaborative problem-solving also smoothed her transition to a professional environment.

Adeesh Kolluru

Adeesh Kolluru ('24) leads an AI team at Radical AI. The company aims to accelerate materials research and development by integrating machine learning methods, computational materials science tools, and autonomous labs. "It's exciting to think about how we can leverage AI to accelerate materials discovery. We see materials everywhere. If we can find any better materials for different applications, we can enable a lot of technologies that we don't currently have," says Kolluru.

Kolluru started building machine learning methods to solve similar problems as a member of John Kitchin's research group. While the group largely focuses on catalysis applications, the methods Kolluru developed for his dissertation also apply across molecules and materials, mapping to his work now.

Internships at Samsung Semiconductor, Orbital Materials, and Meta also connected Kolluru to the field of AI for materials discovery. "It's a very new field, and that means we should use new ways to think about how we can solve the problems. Startups are in the best position to do that," he says. Working at a startup, Kolluru is ready to go from developing fundamental methods to delivering solutions for customers.

Yuri Sanspeur

Yuri Sanspeur ('24) also works at Radical AI. "To streamline materials discovery, we are building an end-to-end stack, from simulations up until the synthesis of actual compounds," he says. The company's experimental results are fed back into their computational simulations. Sanspeur, an AI research scientist, is modeling how metals deform under various stresses. His work is part of the search for new materials that are tough and can be shaped into different forms.

Sanspeur is building on his Ph.D. research, which was primarily computational modeling of metal oxides. Although he's working with similar tools, the assumptions and nuances around the parameters are different. With his advisors Zachary Ulissi and John Kitchin, Sanspeur also used graph neural networks to approximate the potentials of molecules. "We had access to a lot of new technologies. Our abstract chemistry and theoretical work was always married with very modern tools," says Sanspeur.

At Radical AI, he has access to different software to deploy his calculations. The company is building its own supercomputer. Sanspeur helped set up the infrastructure for running their calculations during his first few months on the job. He credits his advisors' practical approach with preparing him to get up to speed very quickly. "My Ph.D. experience equipped me with all the skills necessary to actually implement and run things," he says.

Brook Wander

Brook Wander ('24) is a computational scientist on the materials discovery team at SandboxAQ. The business-to-business (B2B) company is developing solutions at the intersection of AI and quantum technologies. "We're building a computational platform for high-throughput screening and design of novel catalysts, aimed at identifying materials with transformative potential for industrial and energy applications," says Wander.

Catalysts are central to the production of fuels, fertilizers, pharmaceuticals, and clean energy technologies. Improving or discovering new catalysts can have enormous environmental and economic impact. Wander and her team are developing new datasets and leveraging existing public datasets to train models that are capable of capturing the complex physics and chemistry underlying catalytic reactions.

Her work at SandboxAQ aligns directly with experience from her Ph.D. program. In both her doctoral research with advisor John Kitchin and an internship at Meta, Wander applied computational and machine learning methods to understand catalytic materials. "As a chemical engineer, I am trained to connect molecular-level mechanisms with macroscopic performance, which is exactly the kind of reasoning required in computational catalysis," she says. "While the tools have evolved, the way of thinking and the scientific intuition that chemical engineering fosters remain central to what I do every day."