Work

04 projects

  1. 01 Multi-view pedestrian detection with partial labels 2026 Computer vision MVDet MODA 22.7 → 81.5 with 60 % of labels dropped (Wildtrack) PythonPyTorchCUDAOpenCVNumPyCamera calibrationBEV projectionKalman trackingUnityC#DockerYOLO
  2. 02 Face recognition attendance 2025 Computer vision +5 % accuracy vs YOLOv5 on real cameras, 10 % less compute PythonPyTorchONNXOpenCVFastAPISCRFDArcFaceNext.js
  3. 03 VNASelf, a multi-agent assistant 2025 Language Supervisor + 3 specialist agents on LangGraph PythonLangGraphLangChainOpenAI APIRAGFastAPIPostgreSQLSupabaseReactDockerPaddleOCR
  4. 04 VieCareer, an AI mock-interview platform 2026 Language Live at viecareer.site · voice interviews over WebRTC and WebSocket PythonFastAPIOpenAI APIOpenAI RealtimeWebRTCWebSocketPostgreSQLSupabaseRedisDockerReactTailwind CSS