Matt Kerkstra

#1 contributor of 45+ engineers across 5 repositories. Operating at Staff scope.

• ›Architected and shipped a production AI clinical-note templating system from empty repo to production in 16 days across 3 services. Graph-based LLM pipeline with Postgres-backed checkpointing, semantic section matching via BAAI/bge-m3 embeddings + Milvus, dual-detector PHI anonymization (Presidio + LLM), SSE streaming, and an admin review UI with structured rich-text editing.
• ›Built the production Kubernetes ML platform from an empty repo. Istio ambient mesh, ArgoCD app-of-apps with Kustomize overlays, KServe + vLLM for self-hosted model serving (embeddings, ASR), Langfuse for LLM observability, Milvus for vector search. 13 namespaces / 9 services. Cost stayed flat; new services deploy with a Helm chart and an ArgoCD app. Six engineers now contribute regularly.
• ›Led MongoDB → PostgreSQL/PostGIS migration of a clinical analyses data model (100M+ records); cut heavy queries from >5s to <500ms. Unlocked a multi-year cascade of product capabilities - spatial segmentation storage, per-patient overlays, and the clinical recommendation engine in production today. Own every layer of that chain, from schema through Python algorithms to TypeScript integration across 10+ versions.

  compounding arc

  postgis seed · 3yr cascade

• ›Built an end-to-end voice-to-clinical-note pipeline: offline-resilient browser capture (OPFS, Service Workers), Jotai recording state, BullMQ async processing, transcription, structured LLM summarization, multi-language support. Migrated the service into the ML cluster for direct in-mesh access to inference infra. 60K+ TS LOC across client and API.
• ›Delivered the unified-appointments backend that landed the largest dental services organization (DSO) contract in the United States - vault practice search, dual-mode practice support, optimizations for legacy PMS integrations.

  huddle view

  one query · imaging × pms

• ›Established a Kubeflow-based experimentation framework standardizing data versioning and model promotion across the ML org.
• ›Run the bi-weekly Backend Guild (18+ months) driving cross-team architectural alignment. Drove org-wide adoption of typed query patterns (Kysely + footgun-prevention bots) and AI-assisted developer tooling - first mover on Cursor rules, CLAUDE.md, and MCP integrations a year before mainstream.

• ›Rebuilt Node.js + Stripe billing to support 2.5× YoY revenue growth while keeping payments under 200 ms.
• ›Implemented real-time fraud and sanctions checks that shut down illicit GPU crypto-mining from embargoed regions, reducing chargebacks >50%.

  request filter

  chargebacks −50%

• ›Introduced bundle splitting, CDN routing, and feature flags, shrinking mean deploy time from 15 min to 6 min for 40+ engineers.
• ›Drove Redux → React Query migration, reducing cold-start data fetches 40% and bundle size 20%.

  time to interactive

  redux → react query migration

• ›Converted a 20-year-old version-control system for F&I forms from VB6/SQL to COBOL + Pick BASIC - responsible for distributed delivery, usage tracking & billing, and integration with F&I systems. Assumed lead role two weeks after onboarding and delivered on schedule.

  FORM-VCS v2.3

  vb6 → cobol + pick basic