Check out my Portfolio Website

👾 I’m interested in ... Mobile Robotics, Autonomous Driving, Robot Learning, and Computer Vision
🛠 I’m currently pursuing a Master's in Robotics at the University of Maryland, College Park
🤝 I’m looking to collaborate on ... projects in AI, robotics, and embodied intelligence

🔬 Currently working at ...

• GAMMA Lab (UMD) — Research Assistant, working on multi-modal perception (camera + LiDAR), behavior planning, and navigation for curbside autonomous robots using MPPI planners, SLAM, and ROS 2

🌟 I have worked at ...

• Symbotic LLC — Perception Engineer (Ex-Intern), developed and deployed Vision Transformer–based perception systems for warehouse AMRs, optimized ROS 2 pipelines on Jetson using TensorRT, and reduced inference latency by 70%

• Bio-Imaging and Machine Vision (BMV) Lab, UMD — Research Assistant, built ROS + OpenCV vision pipelines for crab and mushroom detection, reducing manual intervention by 75%, and applied imitation learning for UR5e robotic arm manipulation

• University of Alberta (MITACS Globalink) — Reinforcement Learning Research Intern, implemented multi-agent reinforcement learning for high-density autonomous vehicle merging, achieving 30% faster responses and 40% fewer collisions in simulation

📬 How to reach me ... kunjgolwala13@gmail.com | kgolwala@umd.edu

🎥 See my latest work in robotics here ->
TMI: OpenVLA Fine-Tuning | Frenet Trajectory Planner | VisionNav | Human-Following Robot

• TMI (Too Much Info): OpenVLA Fine-Tuning
  🔹 Fine-tuned OpenVLA using behavior cloning on LIBERO (MuJoCo) with vision + language-conditioned actions
  🔹 Improved manipulation success by 35% using LoRA fine-tuning and LLM-based instruction filtering

• Frenet Optimal Trajectory Planner
  🔹 Implemented a Frenet-frame planner for lane keeping, lane changes, and vehicle following in CARLA
  🔹 Generated smooth trajectories using quintic polynomials with feasibility checks on curvature and jerk

• VisionNav: Autonomous Navigation System
  🔹 Built a ROS 2 vision-based navigation stack with ArUco markers and EKF sensor fusion
  🔹 Integrated LiDAR (LIO-SAM), VIO, and optical-flow obstacle detection for real-time autonomy

• AI-Powered Human-Following Robot
  🔹 Developed a ROS 2 human-following pipeline using camera-based tracking and SLAM
  🔹 Integrated ORB-SLAM, ICP scan matching, and potential-field obstacle avoidance

• To contribute to impactful projects in autonomous systems and robot learning
• To bridge classical robotics and foundation models for real-world deployment

Thanks for stopping by! Feel free to reach out or collaborate! 😊