LBOS AI
A locally hosted AI assistant focused on speed, privacy, and offline use. Custom inference pipeline, persistent memory layer, and a UI that feels like talking to a real assistant — not a chatbot demo.
I'm Lihan — a student developer from New Zealand who builds software, AI systems, and robots because the best way to understand technology is to make it.
This page is a deeper look at who I am, what I work on, and how I think about building things that last.
A snapshot of the languages, frameworks, and tools I've spent serious time with. Not exhaustive — just the ones that have actually shipped something.
Selected projects from the last few years. Each one taught me something I couldn't have learned from a tutorial.
A locally hosted AI assistant focused on speed, privacy, and offline use. Custom inference pipeline, persistent memory layer, and a UI that feels like talking to a real assistant — not a chatbot demo.
A disaster response platform that helps communities report emergencies and coordinate support. Live mapping, status updates, and resource matching — built end-to-end.
Robot design, programming, and strategy for FIRST Tech Challenge. Tied for second at the New Zealand National Championship — and a real education in building under pressure.
Milestones from the past few years — not a complete story, just the highlights that changed how I think.
Started programming and joined FIRST LEGO League. Built my first robot, wrote my first real bug, and got hooked on making things work.
Our FLL team qualified for the New Zealand National Championship. First time I built something with a real deadline attached to it.
Started coaching younger FLL teams. Explaining things turned out to be the fastest way to actually understand them.
Stepped up to FIRST Tech Challenge. Built a full competition robot, qualified for Nationals, tied for second place.
Shipped my first serious software project — a private, local AI assistant. Also started CrisisConnect as a real-world response tool.
Splitting time between software, hardware experiments, mentoring FLL teams, and figuring out what's worth building next.
A few principles I keep coming back to. Not rules — just things that have worked when I actually followed them.
Reading docs gets you 60%. Building the thing gets you the rest. Every project I start is partly an excuse to learn how it actually works.
Software should respect the people using it. Local-first, minimal data collection, no dark patterns. If a tool can't function offline, that's a red flag.
The best projects solve problems people actually have. I gravitate toward tools that help — disaster response, accessibility, education — over pure novelty.
A live snapshot of what's on my desk right now.
I'm always up for discussing projects, ideas, or interesting problems.