I build things that work at scale and I still ship the code myself. Thirteen years across banking, insurance, payments and gaming: from backend engineer to a CTO running 110 engineers through a company sale, to leading engineering at a £10bn lending bank. Now building Image Vault, an AI likeness‑protection platform for actors, and taking on a small number of consulting and board roles.
OakNorth lending platform
unlocked
Policy Expert
in twelve months
2022
Image Vault Deep Scan
01Image Vault
A secure biometric likeness vault for actors. Talent store terabyte‑scale 3D scan packages and license them to productions through dual‑custody 2FA, an append‑only compliance ledger and a watermarked render bridge into Unreal and Maya. Deep Scan then sweeps nine public platforms for deepfakes and unauthorised use, adjudicating candidates with the vault's own reference set, and runs the takedown pipeline end to end. Live at imagevault.ai.
02Lending platform
Replaced an expensive external "bank‑in‑a‑box" with an internal origination platform for a bank capitalised to £10bn. Loan origination fell from 110 days to 48, and the system carried lending volume from £2bn through £5bn towards £10bn. Full technical strategy for the lending business, board intelligence and the engineering org.
03Agentic spreading
Multi‑agent flows that ingest customer data packs, working directly on PDF and Excel source material, to assess solvency, performance and the future trend of held loans. The first agentic data‑parsing service at the bank, cutting monitoring‑analyst workload by a quarter and putting credit intelligence on tap. Loan documents vectorised so any deal can be interrogated through RAG.
04Knowledge graph & headless compute
A semantic knowledge graph of the bank's services and data stores, exposed through MCP so agents get headless access to internal web apps that were never designed for them. The foundation of a new headless compute strategy: retrofit, don't rewrite. Includes a production admin server with per‑call 2FA on every mutating tool, full audit logging and a curated semantic layer.
05AI design engine
For a London interiors studio raising investment: a bathroom‑design engine exposed as an MCP server, where a customer's preferences become a well‑formed floor plan and asset list, generated right first time. Derived geometry with commands as the only write path, advisory rules for the model and deterministic constraints the engine enforces, and a console the design team uses to own inventory, rules and templates without engineers.
06Generative imagery pipeline
Working alongside a creative director, using image‑generation models in production to augment and extend editorial rushes: generative models wired into a working studio pipeline rather than a demo. Creative programming is where my engineering and the film world meet, and it's the work I enjoy most.
07Real‑time fraud scoring
Sub‑400ms fraud scores on every transaction through Judopay's payment pipeline. A kappa architecture on Kafka pulled and windowed data from walled subsystems so the in‑flight request did almost no work, with Python models trained on transaction history hosted in a dockerised AWS environment to supplement scoring.
08Public payments SDK
Built Judopay's open‑source mobile SDK for other developers, working with platform holders to grow their ecosystems, and was invited to speak at Xamarin Evolve, Xamarin's international conference. Developer relations learned the practical way: by shipping the thing people integrate.
OakNorthDirector of Engineering · Banking
Complete technical strategy for the lending business of a bank capitalised to £10bn. Built the internal lending platform, brought applied AI into the bank, and worked directly with the CEO and senior leadership on board intelligence and engineering performance.
CoinTrackerEngineering Lead, Data & Infrastructure · Crypto tax
Built an engineering domain from the ground up: migrated production compute from Heroku to GCP, delivered the first company‑wide data warehouse and visualisation layer, re‑architected the database access layer, doubled the team, hired a specialist security engineer and took the company through SOC 2.
Policy ExpertCTO · Insurtech
Joined as DevOps tech principal, promoted to head of engineering, then CTO. Grew tech from 50 to 110 people across 12 teams, took deployments from 55 to 350+ a month, attrition from 60% to 5%, contractor reliance from 40% to 10%, and hit 99.9998% uptime. Put machine learning into production with a modern MLOps pipeline. Promoted ten technical managers. Led a successful due diligence for the sale of the company on the merits of the tech department.
Genius SportsDevOps Lead · iGaming
Took a 100% data‑centre company into AWS. Rolled out EKS and Vault, a unified containerised build and deployment pipeline, multi‑tenanted AWS accounts and a security overhaul, setting the standards that flowed out from London to the rest of the group.
JudopaySenior Security & Data Engineer · Fintech
Core payments, the public SDK team and the fraud team. Real‑time risk scoring on Kafka, an open‑source mobile SDK, and a speaking slot at Xamarin Evolve.
Dome TechnologySoftware Engineer · Construction
Mobile and backend engineering, including a mobile defect‑snagging application for construction sites.
Queen's University BelfastBEng Computer Science · First‑class honours
Year in industry in big data at HP Autonomy. Dissertation on sentiment analysis of the Twitter firehose using distributed computing.
Startup building
Zero to one, fast. Fractional CTO for founders who need a product that works, an architecture that won't need throwing away, and a demo that gets a round done.
- Fractional CTO / first technical hire
- Investor‑ready MVPs
- Hiring the first engineering team
- Architecture that scales from day one
Creative programming
Engineering for studios, directors and creative teams. Generative models in real production pipelines, likeness and image tooling, and the glue that makes new tools behave on set and in post.
- Generative image & video pipelines
- Film, TV & VFX tooling
- Likeness protection & detection
- Interactive and computational work
Applied AI
Agents that do real work inside real companies. I've shipped multi‑agent systems, MCP servers and knowledge graphs in a regulated bank, with the audit, 2FA and security a board will sign off.
- Agentic workflows & MCP
- RAG, knowledge graphs, semantic layers
- AI strategy for existing products
- Production ML & MLOps
NED & advisory
Board‑level technology intelligence from someone who has run the executive side at scale: strategy, budget, hiring, security, board papers and a company sale on the strength of the tech.
- Non‑executive director roles
- Technical due diligence
- Scaling 50 → 110 and beyond
- CTO coaching & succession
I started as an application developer and never really stopped. Backend systems at scale, public SDKs, iOS and Android apps, cloud infrastructure, DevOps culture, then teams, then an org of 110. You name the type of programming, I've done it, and the holistic view is what I bring to a board or a founder.
What I care about is that technology is a value centre, not a cost centre. At Policy Expert that meant taking the company from tenth to sixth in UK market share and through a successful sale. At OakNorth it meant a lending platform that tripled what the bank could do, and the first real AI agents inside it.
Outside work I'm building Image Vault with people from film and television, which is where my engineering and the creative world overlap most. I attend AWS re:Invent whenever I can, read everything, and I'm never satisfied to stand still.
View LinkedIn profile ↗C# · TypeScript · Python · Java · React · Node · Postgres · Kafka · Redis · AWS · GCP · Cloudflare · Kubernetes · Terraform · Vault · GitHub Actions · Supabase · Claude · MCP