Primary
Staff AI Engineer
Production LLM product ownership spanning evaluation, retrieval, model integration, backend services, rollout, and user feedback.
↳ /ai-engineer · role profile
production AI, end to end.
I am a staff-level AI engineer who works at the boundary between AI product, evaluation, model-serving infrastructure, and backend systems. My current domain is clinical AI; the reusable specialty is shipping reliable LLM products whose quality is measured by what users actually do.
[01]
Role fit
Primary
Staff Applied AI Engineer
Turns ambiguous workflow problems into measurable AI systems with human calibration and behavioral feedback loops.
Strong adjacent
Staff AI Platform Engineer
Owns Kubernetes model serving, GitOps delivery, vector retrieval, observability, and a shared platform used by six engineers.
[02]
Evidence
AI product delivery
Shipped a production LLM system with two teammates in 16 days across three services, including checkpointed orchestration, retrieval, privacy controls, streaming, and review UI.
Evaluation and feedback
Combined deterministic diffs, LLM-as-judge, clinician calibration, shadow evaluation, and attributed edit signals; cut edit rate per section from roughly 60% to 40%.
Inference and platform
Owns a Kubernetes ML platform with KServe, vLLM, Langfuse, Milvus, Istio, and ArgoCD across 13 namespaces and nine services while keeping cost flat.
Backend and data depth
Led a 100M+ record MongoDB-to-PostgreSQL/PostGIS migration that reduced heavy queries from more than five seconds to under 500 milliseconds; owns the chain from schemas through Python algorithms to TypeScript integrations.
Full-stack AI delivery
Built an offline-resilient React capture flow, asynchronous TypeScript processing, transcription, LLM summarization, and practice-system integration across client and API.
Staff-level influence
Runs the bi-weekly Backend Guild, authors ADRs, established org-wide patterns, and enables six regular contributors without relying on a management title.
[03]
Boundaries
- The evidence supports production AI systems, evaluation, retrieval, serving, reliability, and product integration.
- The resume does not claim foundation-model research, pretraining, fine-tuning, CUDA or Triton kernel work, or people management.
- Clinical AI is the current proof domain, not a restriction on target industry or role family.