● Mechanical engineering (B.S.) + five years of Unity/C# software (mobile games, 80+ WebGL science sims).
● Robotics is where the two meet — so now I write software that runs on real robots.
🎓 Physical AI Semi-Humanoid Engineer Program, Dongguk AI Campus (2026.07.06–11.04, in progress) · B.S. Mech. Eng., Tech Univ. of Korea
🚗 JD-AMR Cube: ESP32 firmware, C++ ROS 2 driver, Cartographer & Nav2. Completed an approximately 77 m corridor round trip; an obstacle run reached 20 goals with six protective stops followed by same-goal resumption, checked against MCAP logs.
🦾 SO-101 arm: analytical IK and wrist-camera alignment completed pick-and-place. ACT produced one observed autonomous descent and grasp on the same workbench; repeated success rate and retraining effects remain unverified.
🧪 Gazebo: reconstructed the saved map and obstacle scenes as a 2.5D test environment, then tested protective-stop and recovery transitions under synthetic low friction. This is not a live-synchronized or measured 3D reconstruction.
🐾 Isaac Lab / Go2: compared reward conditions under a fixed training budget and evaluated policies separately across directions, terrain, and seeds.
🧪 BreakTok — IMU puck that notices doomscrolling from hand rhythm (TOYTHON 2026, top-20 final)
🌱 Write-ups of everything I build and read: hello, robot
🕹️ Before: Hardboiled Express (Google Play Featured, ★4.7) · Dannect (80+ WebGL science simulations)
| 📍 | Korea · Physical AI program through 2026.11; seeking junior robotics software integration and testing roles |
| 🔧 | Hands-on: ROS 2, C++, Python, Linux, Cartographer, Nav2, MCAP, Gazebo, analytical IK, LeRobot / ACT, MuJoCo, Isaac Lab |
| 🧪 | Current work: AMR obstacle handling and reproducible simulation, SO-101 grasp evaluation, Go2 policy comparison |
| 📚 | Start here: portfolio · AMR code & evaluation · SO-101 · Go2 results |
