Unlearn AI Digital Twins for Clinical Trials

Unlearn.ai, Inc. is a San Francisco company founded by Aaron Smith and Jon Walsh that applies generative AI to clinical development. It addresses the high cost and slow enrolment of clinical trials, where large control arms and uncertain statistical power delay new treatments, especially in neurology.

Unlearn builds AI-generated digital twins that forecast each participant's likely outcome without treatment. These forecasts are used at three stages: in trial design to reduce sample size or control-arm size, during ongoing studies, and after readout to re-analyse results. Its PROCOVA covariate-adjustment method was qualified by the EMA in 2022 for primary analysis of phase 2 and 3 trials with continuous outcomes. TrialPioneer, launched in 2026, is an AI workspace for upstream trial planning. Partners include Acumen Pharmaceuticals, VectorY Therapeutics, SOLA Biosciences, Trace Neuroscience, remynd and the CHDI Foundation.

• Digital twin generators: AI models forecasting each participant's untreated outcome.
• PROCOVA: EMA-qualified covariate adjustment for phase 2 and 3 primary analyses.
• Trial design support: sample size and control-arm reduction at fixed statistical power.
• Synthetic controls: digital twins as benchmarks for single-arm and early-phase studies.
• Post-readout analysis: re-analysis of completed trials to strengthen evidence.
• TrialPioneer: AI workspace for upstream clinical trial planning.

Pharmaceutical sponsors running regional phase 2 and 3 neurology trials with Vietnamese sites can reduce control-arm enrolment. CROs in Vietnam can offer digital twin-based design and analysis services. Academic hospitals can re-analyse investigator-initiated trials to strengthen evidence.

Unlearn reports a 33% smaller control arm and more than 4 months of enrolment time saved in phase 3 trials for a top-5 global sponsor. A phase 1/2 design case cites about 20% total cost savings, roughly 2 million USD. In a single-arm re-analysis, power on Total Functional Capacity rose from 70% with propensity score matching to 86%. EMA qualification of PROCOVA gives sponsors a regulator-accepted method.

Regulators including the EMA and FDA are opening pathways for AI and covariate adjustment in trials, and the FDA has piloted AI-enabled optimisation of early-phase studies. Sponsors running trials in Asia, including Vietnam, face pressure to cut enrolment time and costs, especially in neurodegenerative diseases where patients are hard to recruit.

Unlearn.ai, Inc., 303 2nd Street, San Francisco, California. Founders Aaron Smith and Jon Walsh; CEO Steve Herne. Board includes representatives of 8VC, DCVC Bio and Insight Partners. PROCOVA qualified by the EMA in 2022. Won the Clinical Trial Design category of Fierce's 2026 AI Innovation Awards.