Fractional CTO · AI Systems Architect · 30 years at enterprise scale
I'm not an AI architect who discovered data last year. For three decades I've built data systems a tier-1 investment bank's risk function trusts with billions — multi-terabyte platforms, thousands of regulated reports a day, under full audit. Today I bring that same rigour to AI. I designed and shipped MOTLogs — a live, multi-tenant SaaS built on a voice-AI agent, an agentic RAG knowledge base with its own governance loop, and a bespoke code-intelligence platform exposed to AI agents over MCP.
The gap I close
I'm an AI architect who has carried that whole distance — from retrieval quality and agent design through evaluation, governance, security and the boring infrastructure that keeps it paid for and online. The evidence below isn't a deck; it's a live product, in one person's hands, end to end.
Track record
Before MOTLogs, three decades making data dependable at the scale and regulatory rigour a global bank demands. This is the foundation the AI work is built on — not a pivot away from it.
What the market needs in 2026 — and what I've shipped
The #1 enterprise constraint of 2026 is RAG quality at production scale.
A Wiki knowledge base with hybrid dense + full-text retrieval, a cross-encoder reranker (bge-reranker-base, int8 ONNX), GraphRAG, and grounded citations carrying source URIs into every answer.
Agents are only as good as the tools and context you give them. MCP is the new integration layer.
A production MCP server (mot-ledger) exposing 25+ tools — semantic search, dependency graph walks, 'why does this exist' provenance, invariant guards, refactor planning — that turn a whole codebase into a queryable substrate for AI agents.
Validation and governance are where AI programmes live or die. Eval is the new unit test.
Drift detection, self-evaluation runs, RLHF reweighting, active-learning queues and golden-set ranking validation — wired as hosted services with a governance portal a human reviewer signs off in.
Voice is the highest-bar agent surface: latency, interruptions and reliability are unforgiving.
A Twilio ↔ ElevenLabs ConvAI orchestrator with a custom-LLM contract, spike-driven engineering, and explicit, measured SLA budgets (p95 latency gates) — built on telephony I'd already shipped at PhoneBounce.
Boards now ask 'can we trust the output, and does it survive audit?' before 'is it clever?'
The instinct from 30 years of regulated banking data: multi-tenant row-level security, DSAR/privacy, a guardrail hook that blocks agents answering before they consult the source of truth, and audited overrides.
An architecture that can't be deployed, recovered and paid for is a slide, not a system.
Reproducible Hetzner infrastructure (idempotent bootstrap, Tailscale, disaster-recovery runbook), Stripe billing, nginx/Kestrel, and commit-driven re-indexing — the unglamorous parts that make AI a business.
Flagship · shipped & live
MOTLogs isn't a single app — it's an estate of five production systems I architected and shipped: the voice agent that answers the phone, the self-grading knowledge base behind it, the engine that scores every MOT in the country, the analytics that predict a vehicle's next test, and the code-intelligence brain my AI agents use to build it all. Open any system for the full breakdown.
An AI voice receptionist that answers garage phone lines, books MOTs and routes calls — 24/7.
A multi-channel knowledge base that answers over WhatsApp, SMS and voice — and grades itself.
The log agent: ingests UK MOT test histories and predicts a vehicle's next-test risk.
“The Auction Whisperer” — scores every UK MOT record so auction buyers can read a car in seconds.
A code knowledge graph of the whole estate, exposed to AI agents as 36 MCP tools.
Ventures · Samurai Computing
A multi-tenant platform for UK garages — voice-AI booking, an agentic RAG knowledge base and a code-intelligence MCP. Architected end to end (the flagship above).
Buy phone numbers in 50+ countries and route every call and message to your mobile. Built on Twilio, Plivo, Nexmo and Pusher — the telephony foundation that later became MOTLogs' voice agent.
An ecommerce platform connecting passengers to verified local taxi firms via app, web, SMS and IVR — cloud telephony and geolocation, built and run end to end.
How I work
Thirty years delivering regulated reporting at a global bank means I design for governance, lineage and recovery from day one — the questions a board will ask are answered before they're asked.
Embedder choices A/B-tested, latencies measured live, SLA budgets re-based against reality with a written justification. Decisions leave a trail you can audit.
I orchestrate AI agents to do real engineering — a workflow-driven QA loop ran parallel testers, adversarially verified findings, auto-fixed and redeployed, driving confirmed bugs 13 → 2 to convergence.
Lose the server, run one idempotent bootstrap, you're back. Infrastructure as a runbook, not as tribal knowledge in one person's head.
Stack
Let's talk
Available now for a fractional or permanent CTO / VP Engineering role — or a hands-on engagement to take one stuck AI project to production. Start with the dossier — the one-page version of everything above, ready to forward to your board.
Or call my AI receptionist on 020 3780 8679 — the same voice agent I built into MOTLogs.
Outside work: 8th-degree black belt in kickboxing and a Jiu-Jitsu instructor — I head up three local clubs.
The same AI voice agent Stephen built and ships in MOTLogs. Ask about his experience, the systems on this page, or a role.