I designed and deployed the Invoice Register, a production pipeline that turns PDF invoices into validated structured records. It uses schema-validated LLM extraction, traces latency and cost with LangSmith, and routes low-confidence results for human review.
I'm building a multi-agent web testing SaaS with LangGraph, where agents explore websites, create prioritized test scenarios grounded in a QA knowledge base, and deliver structured reports. The platform includes job retries, idempotency, and per-job token and cost accounting.
My quality engineering work includes automated validation for REST, GraphQL, and gRPC services at Monks, and multi-stack testing for a financial services platform at Banco de Bogotá. I bring that focus on reliability to LLM systems through grounding checks, evaluation harnesses, structured output validation, and production metrics.
