Work

Work & projects

A decade of health-data work: at NIH and Columbia, in pharma and health systems, and at value-based-care companies and startups.

Experience

2025 – present

Data & Analytics Engineer · Sparx

  • Wrote and led the CMS DUA that gave the company its first VRDC access. It was approved by the CMS Privacy Board in May 2026. I defined the study, obtained IRB clearance and drafted the DUA request.
  • Own the rebates pipeline on partner-ACO CCLF claims.
  • Build Part B drug-utilization and provider-comparison tools, with models run inside the VRDC on RIF data.
2023 – 2025

Lead Data Scientist, Value-Based Care · Wellvana (via Signify Health → CVS Health)

  • Led a nationwide VRDC study of Annual Wellness Visits: 100% RIF, 2017–2023, 26–27M fee-for-service beneficiaries per year-pair, with propensity-score IPTW doubly robust regression. AWVs were associated with $923–$1,271 lower total cost of care per beneficiary the following year (~10–12% of average spend), 0.054–0.062 fewer ED visits and 0.158–0.180 fewer inpatient days (all p < 0.01).
  • Rebuilt the company's CMS DUA and added MSSP ACO files the team was eligible for but had never requested.
  • Evaluated post-acute care and chronic care programs for value-based care partners.
2022

Consultant · Sami Health Data

  • Led a small team that built a cloud continuous-ETL system from FHIR endpoints into OMOP CDM v5.4. It delivered 10,000+ patients with ongoing third-party ingestion.
2021

Clinical Data Strategist, Scientific Computing · Mount Sinai Health System

  • Served as the local reference for OMOP and OHDSI. Implemented the OHDSI Data Quality Dashboard and taught live Atlas and Leaf tutorials.
2019 – 2021

Senior Manager, Clinical Informatics · Regeneron Genetics Center

  • Developed electronic phenotypes, both individually and at scale. Appraised new data assets from academic partners and harmonized the in-house ontology with OHDSI.
2018 – 2019

Postdoctoral Research Scientist · Columbia University, Biomedical Informatics

  • Built Chia, an annotated corpus of clinical trial eligibility criteria, published in Scientific Data.
2014 – 2018

Postdoctoral Research Fellow · U.S. National Library of Medicine, NIH

  • Spent 3.5 years as custodian of CMS VRDC Medicare claims: 87M+ patients, 16+ years of follow-up, billions of rows. Recipient of the NLM Special Act or Service Award (2017).

Open-source health-data tools

phea

An R package for electronic phenotyping. You write formulas over long-form patient-event tables and phea generates the SQL, including for values that change over time.

RSQLPhenotyping

atcd

Scrapes the full WHO ATC/DDD index into an analysis-ready table. My most-used open-source tool.

RPythonTerminologies

n2c / ndc_map

Maps FDA National Drug Codes to ATC classes (plus VA classes, MeSH and SNOMED) via RxNorm. It began as an AMIA 2017 paper.

PythonRRxNorm

Medicaid spending maps

Interactive county maps of Medicaid provider spending by HCPCS code and by drug class, 2018–2024, built from 238M claim rows.

PythonMapLibreClaims

VRDC research code

An ETL from the CMS VRDC to the OHDSI CDM, plus SAS study code on Medicare Shared Savings, opioid use, and drugs and dementia risk in the elderly.

SASVRDCOMOP

Beyond health data

Things I build end to end in my own time, from firmware to mobile apps.

Makapix Club

A pixel-art social network built around physical displays, with a web app, an iOS/Android app with an offline editor, and real-time delivery to devices over MQTT. Built solo.

FastAPINext.jsFlutterRust

p3a

A 4-inch touchscreen pixel-art player on the ESP32-P4. It plays Makapix, museum collections and PICO-8, and was featured in Make: magazine.

ESP32C/C++Embedded

p64

A 64×64 RGB LED desk display in a 3D-printed shell, showing pixel art, a clock and the weather.

ESP32-S3C++OpenSCAD

Have a project in mind?

Email me a few lines about your data, your question and your timeline. I reply personally.