◆ Available for freelance & consulting
I build AI systems that have to be right — and prove they are. If you have an LLM feature that works in a demo but not in production, an integration nobody wants to maintain, or a bill that grew faster than the usage, that's the work I do.
Everything I offer below is something you can already see working on this site. No capability listed here is one I can't point at.
Five things, each with a piece of work behind it
Multi-step agents that route, decide and escalate — with a human in the loop wherever a wrong answer would be expensive.
Turning messy documents into structured, validated data your systems can actually rely on.
One standard interface between your internal systems and any AI client, instead of bespoke glue for every tool.
Knowing whether your AI feature works — measured, not assumed — before your users find out for you.
Same output, materially smaller bill — for teams already running LLMs in production and feeling it.
Pick the shape that fits — scope and price agreed before anything starts
Fixed price · about a week
You have something built or half-built and want an honest read: what will break, what it will cost at volume, what to do next. Written findings you can act on.
Fixed price · two to four weeks
One narrow, real use case taken end to end so you can judge feasibility on evidence rather than a demo video.
Scoped project
A production pipeline with evaluation, documentation and a handover session, so your team owns it after I leave.
Day rate
Regular time for architecture review, pairing and unblocking while your team builds the thing themselves.
Half an hour on what you're trying to do and whether I'm the right person. No charge.
What I'll deliver, what it costs, how long, and what I need from you. Agreed before anything starts.
Regular check-ins and working code you can see, not a reveal at the end.
Documentation, evaluation results and a walkthrough — so the work outlives the engagement.
A couple of sentences is plenty — I reply to everything within a couple of days
The ones I get asked most
Agentic system design and build, retrieval-augmented generation (RAG) pipelines, Model Context Protocol (MCP) servers and integrations, LLM evaluation harnesses, and cost optimisation for teams already running LLMs in production.
Yes. Engagements range from a short technical review or proof of concept for a small team, through to building and handing over a production pipeline with an evaluation suite.
Fixed price for scoped pieces of work such as a proof of concept, an evaluation harness or an MCP integration, and a day rate for ongoing consulting. Scope and price are agreed before any work starts.
A Model Context Protocol server exposes your internal systems to AI clients through one standard interface, so an assistant can query your data without bespoke integration code for every tool. This portfolio publishes one you can connect to and try.
By measuring it. That means a golden dataset, an evaluation harness that runs on every change, deterministic verification around model output where correctness matters, and grounding checks so answers can be traced to a source.
Manchester, United Kingdom. Remote work across UK and European time zones, with on-site available in the North West.