Pratham Salvi

Background

About Pratham Salvi

Mechanical Engineering student from VIT Vellore, with hands-on experience applying deep reinforcement learning, computer vision, and control theory to real-world robotics problems. I specialize in building legged robots, autonomous drones, and manipulators—blending mechanical design with cutting-edge AI. At TCS Robotics Research, I’m actively developing bipedal locomotion systems using Isaac Gym and Mujoco, leveraging behavior cloning, imitation learning, and equivariant policies to improve training performance and generalization. I’ve implemented Model Predictive Control (MPC), ZMP control. I'm also exploring video-based humanoid policy learning using vision-based reinforcement learning pipelines. My prior internship at Solar Industries India involved integrating UAVs with YOLOv5-based computer vision systems and simulating robotics workflows in FlexSim to optimize warehouse automation. AI/ML Focus Areas: • Deep Reinforcement Learning (PPO, A2C, BC, Imitation) • Domain Randomization, Sim2Real Transfer • Isaac Gym, Mujoco, PyTorch, OpenAI Gym • Visual Perception using OpenCV, YOLO, SLAM Technical Stack: Python, C++, ROS2, PX4, Gazebo, MAVSDK, Fusion 360, SolidWorks, Ansys Interests: Humanoids, Quadrupeds, AMRs, Manipulators, Autonomous Systems, Human-Robot Interaction