Take generative AI from pilot to governed production.
I help engineering and technology leaders roll out AI coding agents, build agentic RAG and AI platforms, and train their teams — with the security, evaluation and cost controls a regulated enterprise needs.
16 years building and running platforms for
Five ways I help teams ship GenAI that lasts.
Every engagement is built on work I've already done at enterprise scale — not slideware.
AI Coding Agents Rollout
Roll out Claude Code, Codex, Cursor or Windsurf to hundreds or thousands of engineers — with the controls your CISO and risk team will sign off on.
6–10 weeks · pilot to org-wide
GenAI Strategy & Architecture Review
A clear-eyed read on where GenAI will pay off for you, what to build vs. buy, and the reference architecture to get there without rework.
2–4 weeks · fixed scope
Agentic RAG & AI Platform Build
Design and ship knowledge assistants, code-review agents and internal AI platforms that hold up in production — not just in the demo.
8–12 weeks · or as a fractional lead
AI Governance, Security & LLM Cost
Make GenAI auditable, secure and affordable — before the regulator, the incident or the invoice forces the issue.
3–6 weeks
Fractional Head of AI Engineering
Senior AI engineering leadership one to two days a week — strategy, architecture, hiring and delivery oversight — without a full-time executive hire.
Monthly retainer · 3-month minimum
TrainingHands-on workshops and bootcamps
AI coding agents, production RAG, GenAI leadership and secure AI — customised to your stack.
Explore programsAdvice from someone who has carried the pager.
I've done it at scale, in a regulated bank
30,000+ users on AI coding agents and 50,000+ repositories under AI review — with Risk, Cyber and Model Risk signing off.
Engineering and operations first
Sixteen years of SRE, platform and cloud work. I care about evals, telemetry, rollback and cost as much as models.
Vendor-neutral
Anthropic, OpenAI, Google and open-weight models, side by side. Recommendations follow your constraints.
I leave capability behind
Your engineers pair with me, own the runbooks and dashboards, and keep shipping after I've gone.
Outcomes, with numbers attached.
AI coding agents for 30,000+ engineers, inside a regulated bank
Read the story Top-10 US bank (Capital One)An agentic code reviewer across 50,000+ repositories
Read the story Top-10 US bank (Capital One)DeX: an agentic RAG assistant that knows when not to answer
Read the story Paytm$1.8M a month out of a 55PB data platform
Read the storyUpskill your teams on the work they'll actually do.
AI Coding Agents in Practice
Go beyond autocomplete. Engineers learn to delegate real work to Claude Code, Codex or Cursor — safely, reviewably and on your own codebase.
For: Software engineers, tech leads, QA and DevOps
3-day bootcamp · capstone project · up to 20 per cohortBuilding Production RAG & Agents
The engineering behind assistants that answer correctly — and know when not to. Built from lessons shipping DeX across 70+ platforms.
For: ML, platform and backend engineers
Half-day workshop · in person or virtualGenAI for Engineering Leaders
What actually changes in your org when AI writes and reviews code — and how to lead it with numbers, not hype.
For: CTOs, VPs, Directors and engineering managers
1 day · workshop + tabletop exerciseSecure & Governed GenAI
Threat-model LLM apps and agents, instrument them for audit, and rehearse your first AI incident before it happens.
For: Security, GRC, platform and risk teams
Simple to start. Built to hand over.
Discovery call
30 minutes to understand your context, constraints and what success looks like. No pitch deck.
Proposal in 48 hours
A short written scope with outcomes, timeline, team involvement and a fixed or capped price.
Deliver with your team
Weekly check-ins, working artefacts in your repos and tools, and decisions documented as we go.
Hand over, not hand off
Runbooks, dashboards and trained owners so the capability outlives the engagement.
Field notes from enterprise GenAI.
Rolling Out AI Coding Agents to 30,000 Engineers: The Control Plane Nobody Talks About
Getting Claude Code or Codex onto one laptop takes five minutes. Getting it onto thirty thousand, inside a bank, is a platform problem. Here's the control plane that made it possible.
Agentic AI · 7 min readUnder $4 a PR: Designing an AI Code Reviewer Engineers Actually Trust
Most AI review bots get muted within a month. Ours ran across 50,000+ repositories and 350K+ PRs a month. The difference was standards, determinism and evals — not a bigger model.
RAG · 7 min readAgentic RAG That Knows When Not to Answer
The most valuable thing an enterprise assistant can say is "I don't know — here's who does." Lessons from building DeX across 70+ platforms.
Let's talk about where GenAI fits in your organisation.
A free 30-minute call to understand your context. If I'm not the right fit, I'll tell you — and point you to someone who is.