Skip to main content

Projects

Welcome to my project portfolio. Each project here represents a complete journey from concept to deployment, demonstrating the integration of AI, edge computing, and cloud services.

Project Categories​

🤖 AI-Powered Devices​

Interactive systems leveraging cloud AI and edge inference.

🏠 Smart Home & IoT​

Connected devices for home automation and monitoring.

🎨 Interactive Installations​

Art and technology fusion for exhibitions and events.

🦾 Robotics​

Autonomous and semi-autonomous robotic systems.

VoiceButler — Offline Chinese Speech-to-Speech on Orin NX (2026)​

A fully offline wake-word → VAD → SenseVoice ASR → local Gemma → Melo TTS loop on the same Jetson Orin NX 16GB as the rest of the house robot. Speech stays on CPU ONNX so the GPU remains free for Ollama; a keyword router handles driving, arithmetic, and a primary-school textbook tutor before the 2B model ever sees the utterance.

Tech Stack: Jetson Orin NX 16GB, JetPack 6.2, sherpa-onnx, SenseVoice, Silero VAD, Zipformer KWS, MeloTTS, Kokoro, Ollama gemma4:e2b, reSpeaker XVF3800, Roomba

Status: Running on the robot as of August 28, 2026.

PanoTwin — Edge-Side Gaussian-Splatting Digital Twin (2026) · WIP​

A three-stage pipeline running entirely on a Jetson Orin NX 16GB: dual-fisheye calibration, DAP panoramic metric depth, and D²GS sparse-view Gaussian splatting. Walk an Insta360 X5 around a room, get a metric-scale 3D reconstruction — with nothing leaving the LAN. The calibration stage reproduces HKUST's CO-Calib failure analysis to fix intrinsic-initialization divergence on 200°-class fisheye lenses.

Tech Stack: Jetson Orin NX 16GB, JetPack 6.2, TensorRT, CUDA, NVDEC, Insta360 X5, DAP, D²GS, Kalibr, COLMAP, gsplat, ROS 2

Status: Work in progress — code frozen July 28, 2026; the performance metrics are still being re-measured.

How Projects Are Documented​

Each project includes:

  1. Overview - What it does and why it matters
  2. Architecture - System design and component interaction
  3. Hardware - Components and assembly instructions
  4. Software - Code, configurations, and deployment
  5. Lessons Learned - Challenges faced and solutions found

Tags​

Projects are tagged by technology for easy discovery:

  • #AWS - Uses AWS cloud services
  • #Bedrock - Amazon Bedrock AI integration
  • #IoT - Internet of Things
  • #Edge - Edge computing and inference
  • #ESP32 - ESP32 microcontroller
  • #RaspberryPi - Raspberry Pi projects
  • #Jetson - NVIDIA Jetson