Tech Lead · Backend & Platform
I build the platform layer that lets AI agents do real work inside a company.
Most agent demos fall over the moment they meet a real organisation: secrets, tenancy, audit, cost. My work is the unglamorous half that makes them survivable — sandboxed runtimes, a control plane that holds every credential, and guardrails a platform team will actually sign off on.
The shape most of my work takes: nothing sensitive ever reaches the agent.
Multi-tenant agent platform
A platform for running DevOps agents on behalf of many teams at once. Each agent runs as an ephemeral, sandboxed container job and reaches the outside world only through a broker, so the control plane keeps every secret. Review, code, bug-fix and documentation agents coordinate over an event bus.
Azure DevOps · shipped to teams
A webhook-triggered service that reviews pull requests with a real tool-use loop — read the full file, search the codebase, then comment — rather than one shot at a diff. Per-team token budgets keep spend predictable, review history and feedback land in Postgres, and rollout ran in phases so each team could opt in.
Solo build · live
A small business asked for something they could actually run: a public booking flow for clients and a dashboard where the owner sees the day at a glance. Designed, built and deployed end to end, including the parts nobody enjoys — auth, sessions, and what happens when a booking collides.
Happy to talk about agent platforms, backend architecture, or a piece of work you are trying to get off the ground.
Write below and it reaches me directly.
Inbox — visible to you only