



NeuroDiscover AI
Mining evidence for direction
Links
Team
5 members- AL
Alia Merchant
- AM
Amy He
- AY
Ayelet Peres
- KAOwner
Kahini Mehta
- WI
William Yakah
Overview
NeuroDiscover is a multi-agent research engine for pharma translational and commercial strategy teams, as well as academics. It reads across papers, trials, grants, and preprints, finds patient subgroups hidden inside complex diseases like Parkinson's, maps each subgroup to plausible treatment mechanisms, then scores and ranks the resulting opportunities by evidence strength and commercial potential. The output is an auditable, ranked shortlist of subgroup-to-mechanism-to-treatment hypotheses, plus a traceable NIH-style specific aims for each, ready for expert review, partnering, or trial-design exploration. The live demo runs the full six-agent pipeline on Parkinson's: 8 evidence items, 5 subgroups, ranked outputs, and a proposal audit trail with 18 citations. Discovery logic uses public evidence only, with synthetic cohort profiles for visualization, so there is no PHI or HIPAA exposure.