Run the experiment you can't run in real life — grounded in real-world data, in 24 hours.
A new screening program launches. Six months later, uptake is flat in the very ZIP codes it targeted. The post-mortem identifies trust, language, and access barriers — that a synthetic population could have surfaced before a dollar was spent.
The Transfer Center handles more than half of admissions. Black-swan volume surges, staffing changes, and discharge-timing tweaks have non-linear effects on door-to-doctor time. Spreadsheets cannot model them.
A documentation-streamlining initiative reduces clicks for one specialty and triples them for another. By the time the data shows it, the rollout is already system-wide.
Regression on historical data cannot tell you how patients will respond to a program that has never existed. Agent-based simulation can.
Every agent maps to verified U.S. Census and ACS attributes — age, income, household composition, language, employment, healthcare access — down to the ZIP code.
How we think about agent-based modeling, synthetic populations, and the gap between predictive analytics and decision support.
We're running a pre-registered hindcast on the 2021 LA County COVID-19 vaccine rollout — 500 census-grounded LLM personas reacting to real interventions on a simulated feed, blind to the outcomes we'll later compare against.
Simulation Labs is our open-access product — a way for anyone to experience the power of population-scale AI simulation without a sales call.
Three families of decisions where simulation has the highest leverage. Each delivers a clinical-champion-ready report with predicted outcomes, equity stratification, and intervention sensitivities.
Simulate community health worker programs, mailer campaigns, mobile clinics, navigator pathways across real ZIP-code demographics. Surface equity gaps before deployment.
Example: colorectal screening in Boyle Heights — predicted uptake by age × language × insurance over 6 months, under three campaign designs.
Model the Transfer Center, ED, and inpatient flow as agent populations with capacity constraints. Stress-test discharge timing, staffing changes, and surge response.
Example: simulate Keck Hospital's Transfer Center under a 30% flu-season volume surge — bed-block points and Pareto-optimal staffing.
Simulate provider workflows as agent populations. Test inbox-redesign, top-of-license role alignment, automated waitlists — measure friction, retention proxies, and adoption curves.
Example: model a documentation overhaul across 23,000 providers — net administrative burden by specialty, before rollout.
X Research was founded by a USC Keck Medicine staff member working alongside clinicians, operators, and researchers across the system. Our first reference designs — Transfer Center throughput, CBSA equity intervention, provider workflow simulation — are anchored to the operational realities of a high-acuity academic medical center.
X Research is an independent company. References to USC Keck Medicine reflect the founder's individual employment affiliation and do not constitute endorsement by the University of Southern California.
Tell us what intervention you want to test on a synthetic population. We will scope a pilot tailored to your health system.