Manager, Data Science and Engineering · Baltimore Health Analytics
Production analytics in regulated healthcare: CMS Star Ratings, HEDIS, and value-based care.

I build and govern production analytics systems in regulated healthcare: pipelines, semantic layers, and CMS Star Ratings quality-measure systems that clinical, operational, and executive teams can trust and act on.

I started as a news writer and book editor, moved into healthcare research at UW-Madison, and now lead data engineering. That path still shapes how I work: clear questions, defensible methods, and results that change decisions. Today I'm a player-coach manager, building and running the pipelines while managing the engineers who own them.

• Production data pipelines — EHR data modeling across Epic, Cerner, Veradigm, and athenahealth; ETL/ELT on AWS and Databricks; stored procedures built to survive edge cases in regulated populations
• Analytics governance — semantic layers, metric definitions, and dbt-versioned models that give stakeholders one version of truth
• Value-based care analytics — ACO, MSSP, HEDIS, and CMS Medicare Stars performance tracking; benchmarking and significance testing against CMS methodology
• Platform reliability — observability, incident response, and cost optimization across cloud data infrastructure

• Medicare Advantage Insight Engine — A self-running daily feed that separates CMS Medicare Advantage rulemaking signal from press-release noise, scoring each item on six named features and posting what clears the threshold.
  Daily Briefing · Live feed · Source · Methodology post
• Stars Cliff Simulator — An interactive teaching demo of the 4.0 star Quality Bonus Payment cliff in CMS Star Ratings. Vanilla JS, no dependencies, synthetic weights.
  Live demo · Methodology post
• Healthcare Workforce Transition Platform — An O*NET-based model estimating reskilling transition probabilities across the healthcare workforce, sorting targets into Ready Now, Trainable, and Long-Term Reskill with a skill-domain gap analysis.
  GitHub · Post
• ECDS Shock Index — A HEDIS Electronic Clinical Data Systems measure prototype that models how rates shift when ECDS replaces claims-based ED coding, and estimates the Stars cut-point effect at the plan level.
  GitHub · Post
• Care Delivery Workflow Changes — An interrupted time series evaluation of a care-delivery redesign, separating the level shift at implementation from the underlying trend.
  Case study · Post
• Practice Automation Analytics, healthfinch Charlie — A return-on-investment model for clinical practice automation, built on Epic Clarity extracts and delivered as a stakeholder-facing dashboard.
  Post

• Baltimore Health Analytics — Manage two data scientists (a lead and an IC); own measure methodology, pipeline reliability, and release governance for a Medicare Advantage Stars analytics platform. Built a self-service forecast-versus-actual Stars cutpoint dashboard adopted by data science and the CEO.
• Health Catalyst — Led the post-acquisition migration of all analytics to a multi-cloud (AWS + Azure/Databricks) stack. Cut monthly storage cost ~50% and weekly load latency by 24+ hours; standardized SQL into reusable stored procedures (~70% codebase reduction).
• healthfinch — First analytics hire. Built a HIPAA- and HITRUST-compliant platform from scratch unifying app data and four EHR systems; authored Epic Clarity report libraries deployed across 50+ health systems; governed dashboards absorbed 7× user growth and saved 400+ prep hours per quarter.

Long-form essays on healthcare data engineering, Stars methodology, and measurement. A few recent ones:

• My Cap Falls by a Third. My Work Falls by a Sixth.
• BTEQ Still Has a Job
• The Metric Isn't Wrong. It's Just Not Where Quality Lives.
• One API call, three ways to split it: a FRED case study
• Should I Buy RAM Now?

Full archive at zaherkarp.com/blog.

Peer-reviewed work on accountable care (Medicare Shared Savings), clinic design and team efficiency, and EHR optimization in primary care, from my years at UW-Madison.

6 peer-reviewed publications (Google Scholar, ResearchGate). The two most cited:

• Approaches and challenges to optimizing primary care teams' electronic health record usage — Journal of Innovation in Health Informatics (2014)
• Influence of environmental design on team interactions across 3 family medicine clinics — Health Environments Research & Design Journal (2019)

Engineering & data

Cloud & BI

Healthcare

M.P.H., Biostatistics (2015) and B.A., English Literature (2007), University of Wisconsin-Madison. Graduate Certificate in Patient Safety / Human Factors Engineering (2015).

_{Title, stack, writing, projects, and research above are generated from zaherkarp.com's sources of truth (see the colophon). · Madison, WI · Remote}