We spent years inside pharmacovigilance systems at CROs. We know exactly what ROR misses.

Kevin spent seven years building signal detection systems at contract research organizations, supporting post-market safety monitoring for pharmaceutical clients across cardiovascular, oncology, and infectious disease portfolios. The pattern he kept seeing was consistent: by the time the ROR spike appeared in the FAERS data, the clinical team was already reacting to an adverse event cluster that had been accumulating in the record for twelve to eighteen months. The detection was always retrospective. The window for early action had already closed.

The question that led to TrialVyx wasn't "can we make ROR faster?" ROR doesn't have a speed problem — it has a structural problem. It examines drug pairs in isolation. In polypharmacy regimens, the adverse event signal belongs to the combination, not to any individual agent. Pairwise analysis will never surface a triad signal because the triad doesn't exist in its data model.

Kevin began building a graph-based detection model in 2022, working evenings while completing a consulting engagement for a mid-size specialty pharma company's PV department. Priya Mehta — who had been working on biological network analysis before moving into pharmacovigilance informatics — joined as co-founder and CTO. The model produced its first early detection results against a historical FAERS validation corpus in early 2023. TrialVyx, Inc. was incorporated in Boston that spring.