Epidemiological and causal-inference method applied to administrative, surveillance and registry data. Evidence a program office can act on.
Study design, modeling, reporting on large administrative and public-use data. Built for non-technical reviewers.
Linking records with no common identifier. Linkage error quantified, not assumed.
Real effect separated from measurement, linkage and instrument artifact.
Relational schemas, multi-system pipelines, regulated-data foundations.
NLP over case files, clinical literature, technical documentation. Citation locking.
Realistic cohorts where accuracy is measured, not asserted.
Open-weights models behind the client firewall. Controlled data never leaves the boundary.
Formed May 2026. Five completed methods analyses: BRFSS, NSCH, FAERS, IDEA Part B, and synthetic linkage cohorts, public at github.com/UnconfoundLabs. Master's thesis work partnered with CDC / ATSDR. SBIR Phase I submitted 2026 to the Department of Education (IES) and the Air Force.
Replicated methods papers and analyses with responsible AI use. Public at github.com/UnconfoundLabs.
MHS epidemiology, Johns Hopkins. MPH, Emory. Doctoral training in epidemiology and data science.
One accountable technical principal. No subcontractor overhead.
JCP / DD Form 2345 approved August 2026. CMMC Level 2 self-assessment posted to SPRS.