Agentic AI for the Wet Lab

Three talks trace the path from model to bench for people doing computational biology, drug discovery, and applied machine learning. The evening opens at the representation layer, where sequence and structure become embeddings a system can search, then moves up to genome-scale modeling and what post-training buys on real clinical benchmarks, and closes with tooling that turns retrieved and ranked evidence into bench work, scripting, and documentation. Stefano Massaroli (Radical Numerics) presents Omnii, a genome language model pretrained with native fusion of annotation tracks alongside DNA sequence, covering block convolutions and sparse attention over a 2 million base pair context window. Daniel Leuck (Ikayzo) and Calvin Duong (University of Tokyo) use influenza research as a case study for Reveria, a beta platform for AI-assisted wet-lab work built around modular CodeLets. Ewa Szyszka (Qdrant) walks through DrugCLIP, which places molecules and proteins in a shared embedding space so that binding questions become vector search. Doors open at 18:00, talks run 18:30 to 20:00, and networking follows until 21:00. The framing throughout is what is already viable in silico and where wet-lab validation remains the bottleneck.