Hi, I'm Louis

I'm an Artificial Intelligence undergraduate at Xiamen University Malaysia, and I build things because I genuinely can't help it — half my "coursework" ends up turning into a side project I didn't need to build, but wanted to.

A bit more about me

I got into computer science the way a lot of people do — curiosity that turned into obsession. Somewhere along the way I stopped seeing "AI" and "coding" as separate skills to learn and started treating them as tools to solve problems I actually care about: healthcare gaps, career uncertainty, and the small, unglamorous friction points people run into every day.

Outside of code, I play guzheng, stay active (gym, basketball, badminton, running — whatever's on that week), and I'm the person my friends come to when their laptop, their workflow, or their life admin needs debugging.

What I'm good at

• Full-stack problem solving — I don't stick to one layer. Across projects I've touched Android (Kotlin/Jetpack Compose), backend systems (Node.js/TypeScript), and ML pipelines (PyTorch), often on the same project.
• Turning ambiguity into a working prototype, fast. Hackathons are basically my favorite stress test for this — given a vague problem statement and 24-48 hours, I default to building something real rather than just slides.
• Debugging environments, not just code. A weirdly large amount of my skill set is "why won't this build/install/authenticate" — CI issues, dependency hell, API auth quirks. I've learned to enjoy it.
• Making AI genuinely useful, not just impressive. I care more about whether a multi-agent system actually helps a discharged patient remember their medication than whether the demo looks flashy.

What drives me

I like building things that sit close to real consequences — a patient's recovery, a student's career decision, a small business's funding search. It's easy to build a technically impressive project that solves a fake problem; it's much harder (and more interesting to me) to build something modest that solves a real one.

I'm also driven by a low tolerance for "that's just how it's done." A lot of my projects started from noticing a gap — post-discharge care that just... stops, career guidance that doesn't account for Malaysian financial realities, funding platforms that don't talk to each other — and deciding to just build the missing piece instead of waiting for someone else to.

A project I'm proud of: MediSync+

At UMHackathon, my team and I built MediSync+, an AI-powered post-discharge companion app for chronic care patients (starting with Congestive Heart Failure). The idea came from a simple but frustrating observation: the moment a patient leaves the hospital is often the moment the safety net disappears.

We built:

• A Kotlin/Jetpack Compose Android app for medication tracking and symptom monitoring
• A multi-agent AI backend (clinical reasoning, patient data, scheduling, and notification agents working together) instead of one monolithic chatbot
• A secure integration layer connecting the app to a mock hospital backend

I mainly owned the Android build/UI work and squashed a long list of gnarly bugs (JDK mismatches, Compose rendering issues, Hilt/KSP version conflicts) while the team built out the AI layer together. We made it to the finalist round, placing 11th — but more than the placement, it was the first project where I felt like the tech decisions were actually in service of the patient, not the other way around.

Currently exploring

• Deep learning with PyTorch (most recently, a CNN/CNN+LSTM music genre classifier)
• Computer vision fundamentals (stereo vision, image transforms, connected-component labelling)
• Making local/open-source LLMs actually reliable enough for real agentic workflows, not just demos

Let's connect

• [email protected]
• linkedin.com/in/louis-chua-khai-yi
• github.com/chuakhaiyi