AI Platform · MLOps · ML Infrastructure · Production AI
I build production AI systems with an emphasis on reliability, observability, evaluation, and controlled execution.
3+ years of production experience across ML platforms, model serving, Generative AI, and data systems.
- ML Platform Infrastructure — Kubernetes ML control plane with Argo, MLflow, KServe/vLLM, canary serving, and observability · 645K load-test requests, 0 failures
- Agentic SRE — deterministic, evidence-grounded Kubernetes RCA with bounded read-only investigation, causal topology, replayable evidence, and an mTLS connector trust boundary · 35/36 injected causes identified, 0 false strong authority
- Agentic Customer Service Platform — production-oriented agentic AI control plane with RAG, HITL, policy-controlled execution, and deterministic safety boundaries
- Knowledge Base RAG — multilingual hybrid retrieval, reranking, grounding, and evaluation · 95.9% evidence recall
- DecisionSQL — governed Text-to-SQL with deterministic execution boundaries · 88.9% task success
- Agentic Security Audit — security audit skill for tracing how untrusted agent context can reach execution authority across tools, MCP, RAG, memory, and approval flows
- CauseTune — controlled QLoRA fine-tuning with frozen and held-out evaluation · 98.3% blind diagnosis exact match
Core stack: Python · Go · Kubernetes · FastAPI · PostgreSQL · Kafka · MLflow · KServe · Argo Workflows · OpenTelemetry
Selected repositories are pinned below ↓