Featured Projects

Healthcare AI and data platform projects demonstrating expertise in clinical workflows, compliance, and scalable architecture.

Interactive Demo (Sample Data)

AI-powered ICD-10 & CPT suggestions

CDI Agent RAG Platform - Production Healthcare AI

A real GenAI system built for healthcare documentation, coding, and revenue integrity — designed with production-grade architecture for regulated environments. The system is intentionally hybrid: deterministic first, AI-augmented second. Rules where rules are required, AI only where it adds leverage, everything grounded, explainable, and reviewable.

Technologies

PythonLangChainClaude SonnetMedCPT EmbeddingspgvectorBM25RAGFastAPIPostgreSQL

Key Features

  • Medical-Grade Embeddings - MedCPT biomedical embedding model with 768-dimensional vectors optimized for clinical language, ICD-10, CPT, and guidelines
  • Hybrid Retrieval System - Combines semantic search (pgvector), BM25 keyword search, Reciprocal Rank Fusion, and cross-encoder re-ranking for precision
  • Curated Knowledge Bases - ICD-10, CPT, HCPCS, CDI query guidelines, E/M coding rules, DRG/HCC mappings, HEDIS specifications — all versioned and traceable
  • Grounded LLM Generation - Claude Sonnet in strictly grounded mode with system prompts enforcing context-only answers and explicit refusals when data is insufficient
  • Deterministic CDI Pipeline - Entity extraction → Condition recognition → HCC gap detection (v24/v28 RAF) → E/M analysis → CDI query generation with templates first, semantic search second, LLM fallback last
  • Governance & Safety - Prompt/model versioning, knowledge source traceability, rule validation (age, gender, laterality), LLM usage only on top-N candidates

Impact

Production-ready clinical documentation intelligence with full auditability — every recommendation explainable and defensible. Same architectural principles apply to diagnostics, scientific workflows, and regulated AI across healthcare.

Medical Coding API dashboard showing ICD-10 and CPT code search interface with AI-powered suggestions

Medical Coding API

AI-powered ICD-10 & CPT APIs that convert clinical notes into codes. Production-ready REST API with comprehensive documentation and enterprise-grade reliability.

Technologies

PythonNext.jsVector DBAWSPostgres

Key Features

  • Fast Code Search - Search ICD-10 and CPT codes by code or description with fuzzy matching and full-text search
  • AI-Powered Suggestions - Get intelligent code suggestions from clinical text using advanced keyword extraction
  • Secure & Compliant - All API calls encrypted, no PHI stored, HIPAA-ready infrastructure
  • Usage Analytics - Track API usage with detailed logs, real-time statistics, and insights
  • Simple REST API - Clean, well-documented RESTful API with interactive Swagger docs
  • 99.9% Uptime SLA - Production-ready infrastructure with rate limiting, auto-scaling, and 24/7 monitoring

Impact

Enables developers to integrate medical coding intelligence into their healthcare applications with enterprise-grade reliability and security

Project Screenshot

HIPAA FedRAMP Compliant Data Platform

Built a modern data platform from ground up on AWS GovCloud for Defense Health Agency's telehealth operations. Enterprise-grade data infrastructure supporting secure healthcare data processing and analytics for military treatment facilities.

Technologies

AWS GovCloudStep FunctionsLookerRedshiftKinesis FirehoseGlueS3

Key Features

  • Robust Data Pipelines - Automated hourly data processing pipelines for continuous data flow
  • AWS GovCloud Infrastructure - Secure, compliant cloud environment meeting federal requirements
  • Real-time Data Streaming - Kinesis Firehose for high-throughput data ingestion
  • Data Warehouse - Redshift-based analytics platform for population health insights
  • ETL Orchestration - AWS Glue and Step Functions for reliable data transformation workflows
  • Business Intelligence - Looker dashboards for clinical and operational analytics

Impact

Enabled secure, compliant telehealth data operations for Defense Health Agency serving U.S. service members and their families with HIPAA and FedRAMP compliance

Project Screenshot

Modern Telehealth Data Platform - FHIR & Lakehouse Architecture

Transformed fragmented legacy infrastructure into a modern, interoperable data platform for telehealth operations. Built using FHIR standards, Lakehouse, and Medallion architecture to unify data from multiple systems including Salesforce, ADP, patient records, encounters, claims, and clinical programs with real-time streaming and robust governance.

Technologies

GCPBigQueryPub/SubAirflowdbtKafkaLookerFHIR

Key Features

  • FHIR Interoperability - Standards-based healthcare data exchange ensuring semantic interoperability across systems
  • Lakehouse & Medallion Architecture - Modern data architecture combining data lake flexibility with data warehouse performance
  • Multi-Source Integration - Unified platform ingesting business systems (Salesforce, ADP), patient data, encounters, claims, and clinical programs
  • Real-time Data Streaming - Kafka and Pub/Sub for continuous FHIR resource processing and event-driven workflows
  • Modern Data Pipelines - Airflow orchestration with dbt transformations for FHIR resource normalization
  • Data Governance & Compliance - Robust quality controls ensuring HIPAA compliance, data interoperability, and clinical data standards
  • Advanced Analytics - Looker dashboards for population health insights and AI/ML-ready FHIR data models

Impact

Modernized fragmented legacy infrastructure to create an interoperable, scalable telehealth platform supporting data-driven care delivery and AI initiatives

Project Screenshot

HEDIS Outcomes Automation

Automated quality measurement system for healthcare payers to calculate and report HEDIS measures.

Technologies

SnowflakedbtAirflowPython

Key Features

  • Automated measure calculation
  • Gap-in-care identification
  • Provider performance tracking
  • NCQA compliance validation
  • Real-time quality dashboards

Impact

Reduced manual calculation time from weeks to hours, improved measure accuracy to 99.8%

Project Screenshot

GRAX - B2B Data Value Platform for Salesforce

Enterprise data platform that enables organizations to extract Salesforce data and store it securely in their own cloud environments (AWS/GCP/Azure) for advanced analytics, compliance, and AI/ML workflows. Scaled from early stage to serving 60+ enterprise customers.

Technologies

Node.jsGoReactSalesforceAWSGCPAzure

Key Features

  • Multi-Cloud Data Extraction - Secure Salesforce data export to AWS, GCP, or Azure environments
  • Enterprise Data Ownership - Store data in customer-owned cloud infrastructure for full control
  • Advanced Analytics Enablement - Unlock Salesforce data for BI tools and custom analytics
  • AI/ML Workflow Support - Enable machine learning and AI applications on Salesforce data
  • SOC 2 Compliant - Enterprise-grade security and compliance framework
  • Distributed Architecture - Scalable platform supporting 60+ enterprise customers including Dell, Novartis, and Pepsi

Impact

Scaled from 2 to 60+ enterprise customers, enabling data-driven decision making and AI/ML initiatives for Fortune 500 companies