Technology
Phil — the AI scientist
Phil is the agent who runs our lab. It plans routes, sources building blocks, selects conditions from validated presets, schedules plates on POT-2, reads LC/MS results, and decides what to try next. Phil also searches for analogs, profiles molecules, and runs molecular docking and other calculations.
End-to-end planning
From target molecule to ordered building blocks to executed plates, Phil drives the workflow and asks humans only for decisions that need human judgment.
Grounded in our data
Phil's tools cover retrosynthesis, condition recommendation, and LC/MS verdicts, all grounded in data generated on POT-2.
Autonomous, or in the loop
Phil can drive an entire run on its own. When you want to weigh in — on conditions, edge cases, or just to check in — it threads updates and decisions through Slack. The collaboration is optional, not required. Every action is logged with a full audit trail. You can access Phil too, just let us know.
Tightly scoped access
Phil has tightly scoped database permissions, cannot modify customer-facing code, and asks for confirmation before cost-relevant compute.
