Engineering manager at Cezanne HR. I build agentic AI on Rails, and the teams that ship it.
Ten years in software, the last five on Rails. I lead the Cloud, Integration and Learning teams and stay hands-on in the code.
- AI in delivery. Agent pipelines with review limits so their output doesn't swamp reviewers, error triage on a bot, and a like-for-like measure of AI's effect on production error rates.
- Agentic AI in production. Designed and built a customer-facing AI assistant end to end: state machines, tool use, retrieval, tracing and evaluation.
- MCP servers. Built and shipped an external server with OAuth and an OpenAPI spec written for LLM consumption.
- LLMOps. Feedback systems end to end: structured capture, a taxonomy, alerting, links to traces, deduplication.
- Security for AI systems. Prompt injection hardening and threat modelling, domain allowlisting, URL sanitisation, authorisation hardening on agent endpoints.
Ruby on Rails · Python · PostgreSQL · AWS · Docker · MCP · LangGraph · Claude Code



