Traveler...
Bridging low-level hardware constraints and high-level software abstraction. Building everything from Verilog FSMs on an FPGA to offline AI assistants and remote health monitoring tools.
I'm a Computer Engineering undergraduate at NUS, specializing in Advanced Electronics & IoT with a Minor in AI. I build AI systems across the stack, from agentic applications and autonomous runtimes to inference optimization and the hardware underneath them.
My work spans TensorRT-accelerated vision models, AI infrastructure and CI/CD, and Aiko, my personal OS agent where I experiment with memory, context engineering, RAG, and autonomy. I'm particularly interested in making AI systems faster, more reliable, and capable beyond the demo.
NUS Year 2 · CEG · AI Minor · Teaching Assistant (CS2113)
My personal OS agent: a local daemon, mobile app, Telegram bridge, memory system, and tool runtime wired into one loop. It gives me one interface for files, shell, Git, Docker, calendar, web, voice, image generation, and long-running build work. Under the hood: local Ollama by default, OpenRouter when I want speed, plus a 3-lane Gemini CLI software factory. Its mothership layer turns old repos, scripts, and hackathon builds into manifest-backed skills Aiko can call directly.
Mattermost and VS Code, forked and fused into one room where a person and a dozen local CLI agents work on the same codebase. The bottleneck in agentic development is no longer writing code but landing it, so a custom orchestrator watches activity across the channels and the editor and resolves parallel merge conflicts as they happen, instead of leaving a human to referee afterwards. Agents run locally: unlimited, private, and free of a per-token bill.
A cooperative family scenario game about online harm, built on the fact that 84% of Singapore residents meet harmful content but only 8% of bullied youths tell a parent. Every seat is deliberately dealt a different, incomplete view of the same incident and chooses privately; a deterministic engine resolves the outcome and refuses to let anyone outvote the young person it targets. The credits then reconstruct the real case and turn what each role missed into a conversation.
An autonomous care companion for seniors living alone, built on the premise that silence is an input. A seven-phase kernel — wake, think, revalidate, gate, act, receipt, commit — writes an auditable row before each phase may begin, so nothing it does is narration. When a morning check-in goes unanswered it walks a family contact tree on its own; on-device camera CV screens stroke symmetry, mobility and heart rate without a frame ever leaving the phone.
A live, role-aware map of Singapore that gives citizens, responders, and operations teams one shared emergency picture. SOS signals, reports, live public data, dispatch, and grounded AI guidance all flow through a canonical MQTT and Redis state layer.
The Architect's co-pilot. Researches across papers, repos, threads, and the web; outputs cited 22-section technical blueprints with Mermaid C4 diagrams.
Local-first health companion utilizing a 478-point facial mesh and rPPG for edge-computed vital sign monitoring.
Telegram-native shopping agent that turns messy user intent into TinyFish marketplace fan-out, evidence-backed ranking, and a clearer shortlist.
Privacy-focused desktop assistant with offline Ollama IPC orchestration and built-in task management.
Building software for the Matrix-4 autonomous delivery robot. I optimise real-time SAM3 vision-language inference on Jetson Orin AGX with TensorRT, integrating FFmpeg pipelines and YOLOv8 overlays for low-latency deployment. I built an agentic debugging toolchain that automates SSH execution, ROS 2 and CAN telemetry extraction, log parsing and test workflows across deployed robots. And I led outdoor field testing and calibration of the robot — LiDAR SLAM, Nav2 MPPI controller tuning, docking, and MWAN3 routing.
Guiding Year 2 students through Object-Oriented Programming and software engineering principles. I manage weekly tutorials, help teams navigate collaborative development environments, and debug complex Java codebases to ensure solid architectural practices.
I'm actively seeking opportunities in embedded systems, AI integrations, and full-stack development. If you're working on interesting technical challenges, let's connect.