Skip to content
View ericg1212's full-sized avatar

Block or report ericg1212

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ericg1212/README.md

Eric Grynspan  ·  Data Engineer

Healthcare AI & Fintech  |  New York, NY

LinkedIn   Portfolio


8+ years delivering data systems across regulated environments — production pipelines, cloud-native architecture, and compliance-grade testing in healthcare and financial services. Current focus: revenue cycle management, FHIR R4 claims analytics, and AI enrichment layers with structured validation governance.


Portfolio Arc

P2 classifies denials retrospectively  →  P3 adds AI governance  →  P4 prevents denials in real time

Project Domain Status
P2 Denied: Healthcare Claims Intelligence Pipeline RCM · RWE ✅ Live
P3 Trust but Verify: Clinical AI Governance Engine AI Governance ✅ Live
P4 Cleared: Real-Time Prior Authorization Prevention Pipeline Streaming · Denial Prevention 📋 Planned

Projects

Classifies 257K denied claims by root cause — systematic denials vs. documentation failures — where the remediation path differs fundamentally for each.

Stack Synthea FHIR R4 · Python · Snowflake (RAW → staging → mart) · dbt · Dagster
Scale 495K total claims · 51.9% denial rate · $1.2M+ recoverable · 12 dbt models · 83 tests
RWE T2D/CKD cohort · 104 patients · 54.8% metformin utilization

Dual-validation AI governance — LLM enrichment cross-validated by a deterministic rules engine. Every record routes to Gold (trusted) or Review (explainable reason). No black-box outputs.

Stack FHIR R4 · Python · Snowflake · dbt · Dagster · LLM API · Pydantic
Scale 226 patients · 25,958 clinical records · 6 enrichment categories · 6 rules engine domains
Design LLM-as-Judge blind audit · prompt caching · confidence threshold routing

Proprietary AI builders generate a +92.0% Sharpe ratio premium over third-party integrators (Spearman ρ = +0.800, p ≈ 0.005) across 10 major tech stocks — visualized in an interactive Power BI dashboard.

Pipelines 4 production Airflow DAGs — stocks, SEC EDGAR 10-K, FRED macro, analysis
Storage Hive-partitioned S3 data lake · Parquet/Snappy · Glue catalog · serverless Athena
Quality 184 pytest unit tests · moto AWS mocking · GitHub Actions CI/CD · Terraform IaC

Stack

Python SQL Snowflake dbt Apache Airflow Dagster AWS Docker Terraform Power BI GitHub Actions PostgreSQL pandas pytest


Connect

Healthcare Claims Pipeline   Clinical AI Governance Engine   Sharpe Premium Pipeline

Pinned Loading

  1. healthcare-claims-pipeline healthcare-claims-pipeline Public

    HL7 FHIR R4 → OMOP CDM → Snowflake → dbt → Dagster. RCM: classifies 257K denied claims by root cause — systematic vs. documentation failures. RWE: T2D+CKD metformin utilization cohort. 12 dbt model…

    Python 1

  2. ai-healthcare-pipeline ai-healthcare-pipeline Public

    FHIR R4 → Snowflake → Dagster → Claude enrichment + LLM-as-Judge + Rules Engine → dbt Gold/Review routing. Dual-validation AI governance layer — confidence scoring + deterministic cross-validation …

    Python 1

  3. sharpe-premium-pipeline sharpe-premium-pipeline Public

    $650B in AI spend, 10 large-cap stocks, 3 years of data. Proprietary AI builders outperform third-party integrators by 92% on risk-adjusted returns (Spearman ρ=+0.800, p≈0.005). Airflow → S3 → Athe…

    Python 1