Selected work

Real builds, in production. No demos.

Every engagement below is a working system. Client identities are kept private by default. The work shown here is anonymised at the sector level. Named references are available on request under NDA.

01

Prediction markets, finance

Private client, ongoing

An AI trading system running live across thousands of prediction markets.

A manual trader on a prediction-markets platform cannot watch five thousand markets at once. Mispricings close before a human can even notice them. The client needed a system that did the watching, the research and the execution continuously, with enough discipline to stay out of bad trades.

We built a multi-stage architecture: heuristic screening that filters thousands of live markets to the top candidates in under a second; a research layer that pulls news, RSS feeds and live web data to assess the fundamentals; an arbitrage and confidence engine that gates execution; and a live trader with Kelly-sized position management. All of it is test-backed, incrementally shipped and still being extended.

1,842 tests

Passing, continuously run

137 phases

Incrementally shipped

<1s

4,800 markets filtered to top candidates

02

Sports and talent, agency

Private client, 2026

An AI intelligence layer that reads the agency's inbox and surfaces partnership opportunities.

Agents inside a sports and talent firm were manually scanning emails, shared drives and news for partnership signals. They missed relationships. They wasted hours on data entry. The senior people were doing the junior work.

We built a SaaS that ingests email and shared documents nightly, extracts the entities that matter (athletes, brands, relationships, commercial signals), and feeds a daily recommendation surface. A chat agent answers questions against the full knowledge base. A weekly strategic roundup turns the week of signal into a short themed brief. The team reads the system before they read their inbox.

Nightly

Inbox and drive ingest

156 specs

Request-level test coverage

7 tools

Agent-accessible knowledge-base actions

03

Finance and accounting, SME

Private client, 2026

A monthly auto-invoicing bridge that removed the end-of-month admin from a small business.

A small consultancy was spending two to three hours every month turning a time-tracking tool into Xero invoices. Manual, repetitive, error-prone and the one thing standing between the work and getting paid.

We built a bridge that polls the time-tracking API, maps entries to Xero contacts using a rate card, generates draft invoices with attached timesheets, triggers on a fixed cron or on demand, and sends a WhatsApp summary with approve, hold or reject buttons. On approval the invoice authorises and sends automatically. The monthly job now takes a thumb tap.

Hours → seconds

Monthly invoicing effort

20/20

Requirements shipped to v1

Live

Running continuously since launch

04

Luxury hospitality, boutique

Private client, 2025

A branded quote engine and email automation layer that unbottlenecked the founder.

A luxury hospitality business was bottlenecked on its founder. Every quote was being assembled by hand from supplier emails, PDFs and spreadsheets. Every follow-up was being written from scratch. Most of the week went into work that should not have needed a human at all.

We built two systems that sit behind a single admin. The quote engine ingests raw supplier data, applies the firm's pricing and margin rules and produces a brand-consistent proposal in the founder's tone of voice. The email automation layer reads incoming enquiries, drafts tailored responses against the firm's style and follows up at the right cadence. The founder reviews and sends; the system does the rest.

Days → hours

Quote turnaround

100%

Brand-consistent output

1 admin

Quoting plus follow-up, same pane

Want a build like these?

Start with a conversation.

Thirty minutes. No pitch. Bring the part of the business you most wish was running itself.

Book a call