I build enterprise AI and decision intelligence for regulated healthcare. For the last three years I've led the AI and data science function at a regional managed care plan serving 440,000+ members — authoring the enterprise AI strategy, chairing the AI Governance Committee, configuring the GPU infrastructure, and shipping production systems alongside a four-person PhD team. The methodology is operational intelligence deployed at the moment of work: R Shiny and Streamlit instruments where staff, managers, and executives make decisions inside the tool, not outside it. Dashboards summarize trusted data; instruments create it.
My career has been a long arc through regulated environments — USGS, UCLA Center for Health Policy Research, Highmark Blue Cross Blue Shield (where I owned the enterprise R/Hadoop platform from 2017 and wrote the AI/ML ethical guidelines in 2018), L.A. Care Health Plan, Vital Decisions, and now Central California Alliance for Health. The thread is the same in each: build governed analytics and AI capability where the regulatory perimeter is tight, the audience is non-technical, and the work has to compound rather than decay. I think recursively about systems, and I tend to see complete architectures before I build, which is why most of what I deploy is platform first, application second.
Most of my recent production work is internal and lives behind enterprise firewalls, so what's public here is a mix of older experimentation, forks I learned from, and methodology repositories. The first-author publication in Environmental Modelling & Software (2014) on quantile regression and fractional ARIMA for conditional forecasting under non-stationarity is on ResearchGate; a public technical treatment of quantile-based uncertainty quantification is at chronitron.info/post/ts_uncertainty. I read Rumi in the original Persian and find that the recursive, self-referential structure of classical Sufi poetry maps surprisingly well onto how I think about systems architecture — though I keep that connection mostly between me and the work.
Find me: LinkedIn · ResearchGate
Reach me: akbar.esfahani@email.com