Experimental design is a fundamental problem in many science and engineering fields. In this problem, sample efficiency is crucial due to the time, money, and safety costs of conducting experiments in the real-world. In this talk, I will describe a new paradigm for sample-efficient experimental design based on learning in-context generative models for scientific modalities. Through careful studies on initialization, pretraining, and fine-tuning, we show that even in data-scare settings involving non-traditional data modalities, we can learn powerful generative surrogate models that exhibit desirable behaviors for experimental design: few-shot learning via in-context prompting, multi-task generalization with simulated datasets, and online refinement via closed-loop experimentation. Empirically, our generative surrogates significantly outperform long-standing approaches for data-driven experimental design and demonstrate state-of-the-art performance on a range of experimental design benchmarks for physical and life sciences. --------------------------------------------------------------------- ----------------------------------------------------------------- Bio: Aditya Grover is an assistant professor of computer science at UCLA. His research interests are at the intersection of generative modeling and sequential decision making, and grounded in applications for accelerating science and sustainability. Aditya’s research has been recognized with a best paper award (NeurIPS), the Forbes 30 Under 30 List, the AI Researcher of the Year Award by Samsung, the Kavli Fellowship by the US National Academy of Sciences, and the ACM SIGKDD Doctoral Dissertation Award. Aditya received his postdoctoral training at UC Berkeley, PhD from Stanford, and bachelors from IIT Delhi, all in computer science.