I ship AI products on my own.

An AI-native job search intelligence platform. Six agents handle discovery, company research, evidence-grounded CV tailoring, and outcome learning. Every output is grounded in a verified evidence bank, so the system tailors against documented facts rather than generating plausible-sounding claims about work that may or may not have happened.

It refuses to fabricate. If a job requirement has no matching evidence, it flags the gap instead of papering over it. This is not a modest feature. It is the entire premise. Most tools optimise for sending more applications, faster. More slop is still slop. Aplio is built on the opposite assumption: that the signal is in the quality of the match, not the volume of the output.

A professional governance system for PM contractors. Three agents work alongside the user: an Assistant that handles the routine operating cadence, a Challenger that pressure-tests decisions and RAID entries, and a Coach that tracks patterns in judgement over time. It runs the full operating discipline of a contracting career in one place.

It is built to disagree with you. The Challenger is anti-sycophancy AI: when a RAID entry is weak or a decision does not hold up under scrutiny, it says so and keeps saying so until the substance is addressed. Most AI tools are designed to validate and encourage. This one is designed to be right. That is a different thing entirely.

A renovation cost tracker for couples. Budget, wishlist, and receipts in one shared place, so both people are always looking at the same numbers. It answers the question every renovating couple asks every week: where are we against the budget, actually. Not the optimistic version, the real one.

Built for a real renovation, not a hypothetical market. My wife Emma, a psychologist, and I built and used it while we gutted and rebuilt our first home in London. Every feature exists because we needed it that week. There is no padding, no feature invented for a pitch deck.