| Abstract Scope |
Traditional alloy design has long relied on intuition-guided experimentation—an approach increasingly inadequate for navigating today’s vast, high-dimensional compositional spaces. In this talk, I will discuss emerging paradigms in data-driven materials discovery grounded in Bayesian optimization and uncertainty quantification. These approaches enable autonomous, efficient exploration of multi-objective design spaces, integrating information from simulations, experiments, and physics-based models to accelerate discovery while minimizing costly evaluations. I will illustrate these ideas through recent advances in the BIRDSHOT framework, which achieves over 100-fold acceleration in the optimization of compositionally complex alloys. By merging probabilistic reasoning with domain knowledge, Bayesian methods offer a path toward self-driving laboratories for materials discovery and innovation—where alloy design becomes not an art of trial and error, but a science of informed inference. |