Projects

Selected work

Real systems, in production — each one problem → approach → outcome.

FHIR R4 Healthcare Data Platform

  • Python
  • Lambda
  • RDS MySQL
  • Athena
  • QuickSight
  • Medplum
  • LOINC/SNOMED

Problem: ICHOM needed its health-outcomes standard sets published and consumable through a modern healthcare-interoperability API.

Approach: Cloud-native FHIR R4 platform — serverless Lambda APIs with Swagger UI, plus a self-hosted Medplum FHIR server and an automated analytics pipeline replacing manual reporting.

Outcome: Three production APIs live, with end-to-end ETL from synthetic patient data through to QuickSight dashboards.

Real-Time Options Analytics Dashboard

  • Python
  • Dash/Plotly
  • Claude API
  • Market data feeds

Problem: Options-trading decisions needed real-time, multi-source market analytics with dependable automated scoring.

Approach: Live dashboard with Claude-assisted composite scoring, tuned ingestion pipelines, and fallback data providers for fetch reliability.

Outcome: Dependable real-time analytics running in daily personal-production use, tuned for performance and API cost efficiency.

Intelligent Document Automation

  • Lambda (Python/.NET)
  • SQS
  • Aurora MySQL
  • DynamoDB
  • Hyperscience
  • C# WinForms

Problem: High-volume logistics document workflows at Transflo required automated classification and data extraction at scale.

Approach: Event-driven SQS + Lambda pipelines integrating the Hyperscience AI platform, with C# operational tooling for monitoring and administration.

Outcome: Improved accuracy and throughput of production intelligent-automation pipelines processing asynchronous document workloads.

Voice Apps with Composite Scoring

  • Alexa Skills Kit
  • Python
  • Claude API
  • Environmental/tide APIs

Problem: Turning live environmental and tide data into a simple spoken recommendation requires orchestrating multiple APIs and scoring logic.

Approach: Custom Alexa skills with composite scoring over live data feeds, using Claude-designed orchestration for repeatable outputs.

Outcome: Published voice apps delivering consistent, explainable scoring from noisy multi-source data.