Serverless, ML-aware, observable, Terraformed.
I ship working MVPs — four-zone data lakes, multi-tenant RAG, champion/challenger model evaluation — and back the design with real numbers, not vibes. Every project is open source, every benchmark is reproducible, every decision is documented.
A four-zone data lake with event-driven extraction. Benchmarked Amazon Textract against Bedrock Nova Lite on the same canonical contract. The challenger lane absorbed a full model swap — Claude 3 Haiku → Nova Lite — without architectural changes.
A tenant-aware RAG support system on AWS. Hybrid full-text and vector retrieval with strict tenant isolation, JWT-bound query boundaries via Cognito, idempotent ingestion with replacement semantics for living documents.
Multi-tenant AI systems should enforce tenant scope before retrieval. Tenant Lens uses tenant-scoped ingestion, per-tenant indexes, and authorization before data access.
A working AI demo is not the same thing as a promotable workflow. AgentOps Lab explores quality, trace, latency, and cost gates for RAG and agentic systems.
I take on contract and full-time work for serverless platforms, ML/AI integration, and cloud modernization programs. Let's talk.
