Robotics, 3D Vision, and AV Simulation

I design data/vision pipelines and simulation tooling for autonomous systems — from 3D Gaussian Splatting to SLAM to large-scale map/signal curation.

View Projects Get in Touch
Abstract robotics illustration

About

I’m a Robotics & CS student (U-M) with experience across AV simulation and 3D perception. I’ve built WOMD→SUMO toolchains, improved dataset traffic-signal quality at scale, and engineered 3DGS pipelines supporting closed-loop simulation. I care about reliable metrics, clean tooling, and reproducible experiments.

Skills

Languages

  • Python
  • C/C++
  • Bash/Linux
  • CUDA

Robotics

  • ROS 2
  • SLAM
  • 3DGS / NeRF
  • Simulation

AI/ML

  • PyTorch/Lightning
  • HuggingFace
  • OpenCV
  • Open3D

Tools

  • Git
  • Docker
  • CI/CD
  • Datasets/Benchmarking

Projects

All Experiences →
Internship

XPENG Autonomous Driving Center

Engineered 3DGS/Diffusion stack and sped up rendering 4×; standardized Docker.
Research

WOMD→SUMO Simulation Toolchain

0.653 realism on WOSAC’24; long-horizon evaluation for AV simulations.
Data

WOMD Traffic Signal Quality

Imputed signals at scale; cut red-light violations from 15.7% to 1.3%.
Project

3D Pile Modeling for Wheel Loaders

KinectFusion SLAM; ~12–15 Hz updates with < 10 cm error; Alpha-Shape volume.
Research

TPV Semantic Occupancy Ablation

48% mIoU on Panoptic nuScenes with memory-aware plane configs.

Timeline

  1. 2025
    XPENG internship; wheel-loader 3D modeling; preparing for U-M Robotics MS.
  2. 2024
    WOSAC’24 simulation tooling; TPV occupancy ablation on Panoptic nuScenes.
  3. 2023
    Focused on 3D vision and systems; built reproducible ML tooling.

Contact

Prefer email. I’m open to collaborations, internships, and research opportunities.

Email Me GitHub