Resumette

Printer-friendly standard résumé, any HTML tags with web-only CSS class will be hidden on print.

You can click at any sections or lines hide some information before printing.

Panupong Chaichun

AI / Machine Learning Engineer

Information Technology student at KMUTT specializing in AI/ML, Computer Vision, and Data Science, with experience in Software Engineering and Intelligent Robotics.

Education


  • B.Sc. Information Technology, School of Information Technology, King Mongkut's University of Technology Thonburi (KMUTT), Bangkok, Thailand

Work Experience


Freelance Robotics Software Developer
September 2025 – Present
  • Extended Unitree G1 teleoperation for arm and dexterous-hand control using Meta Quest 3 tracking, DDS, inverse kinematics, and real-time image streaming.
  • Adapted Azure Kinect body tracking and SMPL motion processing for humanoid control and simulation via ZeroMQ.
  • Integrated Inspire RH56DFX hands with MediaPipe/TFLite gesture recognition, Modbus RTU control, and force/angle telemetry.
  • Extended Isaac Lab and LeRobot workflows for manipulation datasets, policy training, replay, and simulation evaluation.
  • Integrated FOHEART motion capture with GMR and TeleopIt to generate G1 29-DoF motion datasets for whole-body policy training.
  • Developed a web-based control interface and validation tools for teleoperation configuration and testing.

Technologies and Languages


  • AI / ML: PyTorch, TensorFlow, Scikit-learn, TensorFlow Lite, Pandas, NumPy
  • Computer Vision: OpenCV, MediaPipe, Azure Kinect, OCR and vision-language pipelines
  • Languages: Python, C, C++, JavaScript/TypeScript, SQL, Bash
  • Software: FastAPI, React, Svelte / SvelteKit, Node.js, Vite, SQLite
  • Robotics: Unitree G1, Inspire RH56DFX, Isaac Sim / Isaac Lab, MuJoCo, ROS 2, DDS, LeRobot
  • Tools: Git / GitHub, Docker, Linux, CMake, ZeroMQ

Projects


  • Thai Election OCR Pipeline - Built a vision-language and LLM pipeline to extract structured information from 846 pages across 300 Thai election documents.
  • Unitree G1 Robot Learning Pipeline - Adapted LeRobot workflows for manipulation datasets, policy training, replay, and simulation evaluation on Unitree G1.
  • Inspire Hand Mimic — Vision-Based Robotic Hand Control - Integrated MediaPipe and TFLite gesture recognition with Inspire RH56DFX control, two-hand mapping, and real-time telemetry.
  • Azure Kinect to Unitree G1 Motion Pipeline - Adapted Kinect body tracking and SMPL-based motion processing for humanoid control, with ZeroMQ streaming and replay.
  • XR Teleoperation for Unitree G1 - Extended Unitree's XR teleoperation stack with Quest 3 hand tracking, Inspire dexterous hands, locomotion control, and a web-based operator interface.
  • Unitree G1 + Inspire Simulation - Extended Unitree's Isaac Lab environment to support G1 with Inspire dexterous hands for pick-and-place, manipulation, and simulation validation.
  • FOHEART to Unitree G1 Motion Capture - Integrated FOHEART motion capture with GMR and TeleopIt to retarget human motion into G1 29-DoF trajectories for training datasets.
  • Course Add/Drop Management System - Collaborated on a university enrollment platform using Vite, Express, Prisma, MySQL, Keycloak, Docker, and Nginx.
  • Dev Recall — Developer Memory CLI - Built a local-first Python CLI that indexes shell history, Git commits, and runbooks using SQLite FTS5 for search and weekly digests.
  • Memory OS — Personal Organizer - Built a local FastAPI application for notes, tasks, journaling, authentication, and full-text search using SQLite FTS5.

Achievements & Certifications


  • Super AI Engineer Season 6 — Foundation AI (Theory), AIAT, 2026 - Passed Foundation AI (Theory), covering machine learning, deep learning, computer vision, NLP, RAG, and data science.
  • Kaggle — NVIDIA Nemotron Model Reasoning Challenge - Private leaderboard score 0.848 (rank 666).