Foundation models and agentic AI are opening a new frontier for scientific innovation, one that demands capabilities beyond those built for general-purpose language and consumer applications. Scientific AI must reason across equations, code, simulations, and experimental and observational data; integrate multiple modalities; interact with specialized tools, instruments, and computational systems; and carry out complex, multi-step workflows with high reliability. The frontier is shifting from prediction to agency: systems that plan and execute multi-step workflows, propose and test hypotheses, and close the loop between models, computation, experiments, and real-world data.

#SciFM27 brings together leading companies with scientists, engineers, clinicians, national laboratories, and government to explore advances in scientific foundation models, scientific reasoning and agentic systems, AI-enabled experimentation and therapeutic discovery, and the computational infrastructure needed to build and deploy these systems at scale.

One of the main themes of #SciFM27 will be on therapeutics. Bringing a new medicine from discovery through clinical development and regulatory approval takes 10 to 15 years and costs more than $2 billion, while roughly 90 percent of drug candidates entering clinical trials fail to gain approval — a rate unchanged in 40 years. Recent advances have created opportunities to connect knowledge and data across physics, chemistry, biology, and medicine; generate and evaluate new therapeutic hypotheses; integrate computational and experimental workflows; and improve how potential medicines are designed and tested.