At MBition (Mercedes-Benz), I automated the migration of 70,000 integration tests using MCP and tool-calling agents that converted tests and opened pull requests for engineers to review. I also owned the process for getting models into production, from staged rollouts and regression checks to rollback.
As a freelancer, I build RAG assistants and agent products, including an agent security layer with Docker sandboxes and Rule of Two checks, and a medical intake that gives doctors a structured briefing and ranked ICD-10 shortlist. I care about evaluation, guardrails, and knowing when a model is wrong before a user does.
