Keynote Talks

I will then explore how we can move beyond sequence alone by incorporating structural information through geometric encoder training and inverse folding with structural consistency, both of which enhance a model’s ability to capture the three-dimensional constraints underlying protein function. Finally, I will highlight emerging approaches for modeling protein dynamics using physics-informed techniques, focusing on energy-based alignment as a means to encode conformational landscapes and guide predictions toward physically meaningful states. Together, these developments illustrate how next-generation protein foundation models can unify sequence, structure, and dynamics to accelerate discovery across the biological sciences.

Bio: Dr. Payel Das is a Principal Research Staff Member and Manager at IBM Research AI in New York, with deep expertise across machine learning, physics, biology, chemistry, and neuroscience. A visionary scientist and technology leader, she is dedicated to advancing innovation responsibly and developing scalable AI systems that serve science and society. Her work has driven key advances in trustworthy AI, next-generation generative AI (GenAI), and AI systems for scientific discovery and social good.

At IBM Research, Dr. Das has held several leadership positions, including Research Scientist, Technical Theme Lead, and Chief Scientist. She also serves as Adjunct Associate Professor in the Department of Applied Physics and Applied Mathematics at Columbia University. Previously, she was a Visiting Fellow at UCLA’s Institute of Pure and Applied Mathematics and a Visiting Student at Princeton University. Dr. Das earned her Master’s degree from the Indian Institute of Technology, Chennai, where she conducted research in Density Functional Theory, and her Ph.D. from Rice University in 2007, focusing on modeling protein landscapes through statistical physics and machine learning. As a Postdoctoral Fellow at IBM Research, she applied high-performance computing to free energy perturbation simulations for studying disease mechanisms. Her research aims to accelerate discovery in the natural sciences by developing state-of-the-art chemical and biomolecular foundation AI models that enable faster identification of disease markers, therapies, and materials. She has also designed brain-inspired language models with adaptive memory for improved reasoning and learning, and architected open-source, cloud-based GenAI systems that promote responsible AI democratization. Her work includes developing methodologies for robust length generalization, fine-tuning, and alignment of transformer-based LLMs. Das is also interested in metascience and has proposed "Coopetition" as a mechanism for constructing community-based discovery framework.

Dr. Das holds over 60 peer-reviewed publications and 60 patent disclosures and was named an IBM Master Inventor in 2021. She frequently delivers keynote lectures and serves on editorial boards, program committees, and advisory panels. Collaborating with agencies such as DOE, NIST, and NSF, she has contributed to shaping ethical and safe AI for science. Recognized by the Harvard Belfer Center and IEEE, she is also a strong advocate for diversity in STEM, mentoring early-career professionals and students, especially women, in technology and science.