I'm an active developer and Senior Admin for SiriusMC, a modded Minecraft network running six modpacks. Over the years I've designed, built and maintained a huge amount of the infrastructure that keeps it running — from a from-scratch centralised network API to AI-driven moderation. (All source is private.)

• Built a centralised network API from scratch — the backbone that unlocked nearly everything else.
• Rebuilt the player shop system ground-up with a web dashboard, replacing a limited off-the-shelf plugin.
• Built a Discord AI support bot on OpenAI: it self-manages an internal FAQ, hooks into our ticketing and API, and moderates in-game chat in 20-message context windows.
• Rolled out Auth0 single sign-on across web support, wiki and Redmine — replacing a mess of separate staff logins (some apps needed rewrites).
• Replaced Hesk web support with a ground-up rebuild and a Discord bridge, so web tickets surface in Discord instead of rotting in a spam folder.
• Built a suite of Velocity proxy plugins — SiriusSanctions (a LiteBans rebuild on our API), StaffChat, GroupChat, RoleSync and more — so they run proxy-side instead of on every backend.
• Built and maintain the network's public website, plus internal tooling like the wiki (BookStack) and project boards.
• Migrated the web services (tickets, Redmine, wiki, API) and the AI bot onto Kubernetes.

Heavy ongoing contributions to an experimental, AI-agent-driven decompilation of Star Fox Adventures — my all-time favourite childhood game. Thousands of contributions in, it's been a fascinating way to learn C, the art of matching decompilation, and the wider retro-game decomp community.

A mod for the game Necesse that brings Rimworld-inspired social dynamics to your settlements. Settlers get deterministic personality traits, form asymmetric relationships that drift over time based on proximity and compatibility, fall in love, break up, and have it all feed back into the happiness system.

An automated pipeline that turns Discord chat exports into high-quality FAQ entries. Uses GPT models to identify questions, vector embeddings to cluster similar ones, and contextual analysis to synthesise comprehensive answers across a six-step pipeline.

A daily micro-quest game. Everyone gets the same nine random quests each day — some are simple observations, some take a bit more effort, and some are hilariously impossible, abstract or painfully mundane. You'll only manage a handful at best (by design), so share with friends and see who actually attempts them. Recent quests: “Feel the weight of an essential oil diffuser” and “Hum a song about a necklace.”