Selected products and experiments

What was built

• Android app on Google Play, a web app and the backend behind both.
• A pattern layer that accumulates small moments and surfaces recurring triggers, emotion pairings and movement against a personal baseline.
• Explicit confidence handling, so sparse data reads as not enough yet rather than a false signal.
• Anonymous-first design: the product is useful before anyone hands over an identity.

• Production Android app ·
• Public on Google Play ·
• 100+ downloads ·
• Mobile, web and backend ·
• Shipped end to end

What this demonstrates

• Product formulation
• Behavioural data
• System architecture
• Mobile + web
• Production

A reflection tool, not a clinical one. It reports patterns in what a person logs and nothing beyond that.

What was built

• Fixtures, results and standings retrieved on a schedule and stored, with the capture timestamp shown wherever the data appears.
• Daily bookmaker market snapshots, de-margined per operator and averaged into a market-implied probability, shown beside the product's own model — including where the two disagree and by how much.
• A season simulator that replays all 380 fixtures ten thousand times from their own scoreline distributions: seeded, linkable, and explicit that a pre-season projection carries no information about who is better.
• Alerting against stated thresholds, so a card appears only when something actually crossed one rather than on every change.
• Quota controls, graceful degradation when a provider goes quiet, and published methodology and data-source pages.

• Two leagues on one system ·
• Multiple external providers ·
• Bookmaker market snapshots ·
• Seeded season simulation ·
• Explicit freshness and confidence states

What this demonstrates

• API systems
• Data products
• Prediction logic
• Reliability
• AI-assisted execution

Analysis only. Neither build takes wagers, holds balances, or awards anything that can be exchanged for money.

What was built

• An onboarding flow where the chosen goal shapes the plan and the tone rather than just a label.
• Tracking for alcohol-free nights and money kept, framed as accumulated signal rather than an unbroken run.
• Craving support built for the moment it is needed, not as a retrospective report.
• Local-first storage with a backup path: no account, nothing leaving the device.

• Non-binary progress model ·
• Goal-driven onboarding ·
• Local-first, no account ·
• End-to-end product design

What this demonstrates

• Behavioural product thinking
• Onboarding design
• Privacy-conscious architecture
• Product depth

A self-directed tracking tool. It offers no treatment, diagnosis or clinical advice.

What was built

• A turn-based open world where specification, delivery, inspection and inquiry are modelled as interacting systems.
• Play grounded in real codes and clauses, then bent for satire: M40 specified, M10 delivered, inquiry pending.
• An interface styled as internal municipal record-keeping, so the narrative arrives through the documents themselves.

• Original systems design ·
• Turn-based mechanics ·
• Interactive storytelling ·
• Built on domain experience

What this demonstrates

• Systems thinking
• Game mechanics
• Interactive narrative
• Creative execution

Draws directly on a structural engineering background: the absurdities are specific because the domain is familiar.

What was built

• A multi-chapter interactive story where choices carry through, alongside a rolling bulletin and colony rankings.
• Playful device interaction and share mechanics designed for the way the festival actually spreads.
• Installable to the home screen, with everything processed on-device: no accounts, no tracking.
• Shipped against a fixed cultural date rather than an open-ended roadmap.

• Consumer product experiment ·
• Culturally specific UX ·
• Sharing mechanics ·
• Shipped to a cultural deadline

What this demonstrates

• Consumer experimentation
• Cultural design
• Social mechanics
• Fast execution

What was built

• Live match rooms with presence, reactions and a running feed, spanning 48 teams and 104 matches across three host nations.
• The Oracle: per-match calls from ten thousand simulations, each shown with a confidence score, expected goals, likeliest scorelines and a written read of why.
• A prediction ladder and leaderboard, plus party games — flash predictions, crowd meter, trash talk — built for people already watching.
• Three-step onboarding that sets identity, team and tone, an installable home-screen build, and play-money markets with no cash stake at any point.

• Real-time multiplayer ·
• Explainable AI predictions ·
• 48 teams · 104 matches ·
• Shipped to a fixed global window

What this demonstrates

• Real-time systems
• Applied AI
• Social mechanics
• Consumer product design

Built for the 2026 tournament window, so it now opens on a finished competition. Points and markets are play-money throughout.