At the Massachusetts Institute of Technology, Associate Professor Rafael Gómez-Bombarelli is championing a new era of scientific discovery driven by artificial intelligence and high-fidelity simulations, tells MIT News. With more than a decade of work at the intersection of machine learning and materials science, Gómez-Bombarelli sees the field at a tipping point where computational tools are not just aids but active partners in research. He believes recent advances in generative models, representation learning, and multimodal AI are ushering in what he calls a “second inflection point” in how science is pursued.
Gómez-Bombarelli’s research blends physics-based simulation with machine learning to accelerate the design and understanding of new materials. By using algorithms that learn from data generated by simulations of atomic behavior, his group can predict material properties and guide experiments more efficiently than traditional trial-and-error methods. This strategy has already yielded new candidates for batteries, catalysts, plastics, and organic light-emitting diodes, narrowing the gap between theoretical predictions and practical application.
A central feature of his approach is the virtuous cycle between simulations and AI. Physics-based models, which compute how atoms interact based on fundamental laws, provide high-quality data that trains machine learning models more effectively. In turn, the trained AI models can explore vast material compositions and structures at speeds far beyond manual exploration, guiding researchers toward promising directions and reducing the time from concept to usable material.
Gómez-Bombarelli has also co-founded companies and served on scientific advisory boards across sectors ranging from drug discovery to robotics, demonstrating how foundational AI tools developed in academia can translate into commercial impact. His work signals a future in which complex scientific problems, once deemed too slow or costly to tackle at scale, are approached with computational pipelines that cut months or years from research cycles and open new possibilities for sustainable energy, advanced electronics, and beyond.