Work

• Designed the platform's first public API and the org-wide standard three services now publish against
• Shipped an MCP server exposing platform tools to LLM agents, with an OAuth broker and encrypted token store
• Architected infrastructure enabling customer-owned AWS accounts as isolated/integrated compute planes
• Engineered the access control service governing Snowflake, AWS IAM, and Auth0 across 17 customer environments, with batched asynchronous permission propagation and dead-letter redrive
• Delivered cloud-cost attribution from design to production in 3 weeks, tagging provisioned resources to funding sources
• Decomposed a monolithic Terraform state into 7 dependency-ordered layers, migrated live with no downtime
• Led major-version migrations of the metadata catalog and identity tooling across all customer environments, de-risked in parallel environments before production cutover
• Shipped a cross-platform CLI and shared API client libraries, retiring hand-rolled HTTP calls across 8 services
• Rolled out a feature-flag and progressive-rollout system (Flipt) for gradual releases across services
• Lead a 3-engineer pod owning the public API and agent-tooling roadmap
• Co-authored the platform engineering technical interview and the code and design review standards the team follows

• Built a serverless patient-records ingestion pipeline on AWS Lambda and SQS
• Launched cohort exploration and document search products backed by OpenSearch and a normalized PostgreSQL schema

• Built a production-grade medical-notes NLP summarization system
• Designed an ML training pipeline producing multiple model variants daily on large medical datasets
• Led technical discovery for enterprise AI, scoping a medical-device chatbot engagement

• Developed a real-time ML system monitoring 1,000+ infrastructure points with 90% detection accuracy
• Built physics-guided neural networks improving structural defect prediction by 40% over baseline
• Secured a $750K grant by developing scalable ML infrastructure enabling team expansion
• Published novel ML monitoring techniques at TriDurLE Symposium and deployed to CDOT production