Exploring Innovations at the BO Hackathon for Chemistry & Materials

The BO Hackathon for Chemistry and Materials was a two-day virtual gathering on March 27–28, 2024, organized by the Acceleration Consortium at the University of Toronto in partnership with Merck KGaA. Researchers and practitioners from around the globe came together to push the boundaries of Bayesian optimization in chemistry and materials science.

Over the weekend, 44 teams (≈120 participants) tackled challenges spanning multi-objective benchmarking, multi-fidelity modeling, novel acquisition strategies, and even GPT-powered optimization frameworks. Attendees collaborated in a lively Gather Town environment, enjoying virtual “hallway” chats and focused breakout sessions despite the international time zones.

All project write-ups, code repositories, slides, and demo videos are openly available on the hackathon’s Projects page. Highlights include:

• Multi-objective Benchmarking of Dragonfly vs. BoTorch

• Take Your Time: Improving Optimization with Random Retries

• Fuzzy Domain Knowledge in LIFT-Powered LLMs

Whether you’re looking for state-of-the-art benchmarks, hands-on tutorials, or inspiration for your next optimization challenge, the BO Hackathon site is a rich resource. Dive in and explore these innovations firsthand!

link: Hackathon projects in details