Applied AI prototypes
Zero-to-one prototypes that de-risk an AI product before you commit a team to it. I design, build and demo the thing — and shape the business case around it.
I'm Huzi Patel. A solution architect and applied-AI practitioner working at the point where ambitious prototypes meet real procurement, real governance, and real users. Currently IoT tech lead at Westminster City Council. Available for selected advisory and delivery engagements.
£2–4m / year in missed business rates, surfaced by a vision-language model over Google Street View.
A cross-council computer-vision platform for waste operations, built on the back of cross-functional buy-in.
£200k+ in commercial recovery by breaking supplier lock-in on the housing LoRaWAN gateway estate.
Technical architecture lead for a Mayor-of-London-backed city-wide public WiFi programme.
End-to-end Smart City architecture that supported a winning RFP for a 150,000+ LoRaWAN water-meter rollout.
Real-time, serverless oil-condition monitoring for a multinational miner across African sites.
Engagements are kept deliberately small — one lead practitioner, no account layer. Most work is fixed-scope: a prototype, a review, a target architecture. Longer advisory retainers for programmes that need a steady technical voice over quarters.
Zero-to-one prototypes that de-risk an AI product before you commit a team to it. I design, build and demo the thing — and shape the business case around it.
End-to-end architecture for IoT, data and AI systems. C4 diagrams, threat modelling, supplier selection, governance — the full stack a serious programme needs.
Device-to-cloud design for LoRaWAN / 5G / Edge AI programmes. I have lived inside RFPs for water, waste and public realm — I know what makes them land.
Fractional architect / tech lead on retainer. Second opinion on a programme in trouble, a review of your AI roadmap, or a credible voice in a supplier negotiation.
Discovery with the people doing the work, not the people writing the brief. The business case is usually hiding somewhere else.
A rough, working prototype resolves ten arguments cheaper than a steering committee. I build early, demo often.
Design for the team that will inherit the system, not the one that built it. Governance, documentation, and a honest view of risk.
An AI prototype that never reaches production is an expensive demo. I stay in long enough that the thing survives me.
Huzi walked in with a prototype, a business case and a working relationship with the team that needed it. We'd been trying to get that initiative off the ground for two years.
He's the rare architect who's as comfortable in front of a CTO as he is pair-programming on a Friday. That's what made the LoRaWAN recovery actually land.
On a 150,000-meter RFP you need a solution architect who can hold the full stack in their head. Huzi could — and the response showed it.
// quotes are representative and anonymised; named references on request
MEng Chemical with Nuclear Engineering, Imperial College London (2:1, 66%). McNab Lacey Prize — 1st place out of all UK universities for designing an innovative process converting waste wood into jet fuel.
A-Levels: Mathematics (A*), Chemistry (A*), Biology (A*), Physics (A) — Canon Slade Sixth Form, Bolton.
Languages: English (native), Gujarati (fluent), Hindi (fluent).
Applied AI, IoT platforms, public-sector delivery, or getting a messy AI prototype across the line to production. Advisory, fractional, or fixed-scope — happy to chat.