I'm a robotics and machine learning engineer at Luminous Robotics in Boston. I build perception and learning systems for robots: vision transformers, 6-DoF pose estimation, Isaac Sim digital twins, and the hardware they run on.
- Now: building digital twins in NVIDIA Isaac Sim and fine-tuning pose estimators (FoundationPose, CenterPose, DOPE) as part of the AWS × NVIDIA Physical AI program.
- Before: ML Scientist at Oxipital AI, working on panoptic segmentation, self-supervised ViT backbones, and cutting TensorRT inference latency from 300 to ~110 ms.
- On the side: a 3D-printed T-rex biped that learns to walk in simulation and now walks on real hardware.
- Education: M.S. Mechanical Engineering (Robotics & Control), Columbia University; B.E., BITS Pilani.
Toolkit: PyTorch · ONNX / TensorRT · Isaac Sim / Isaac Lab · ROS 2 · MuJoCo · Docker · AWS
- walking_trex_project: sim-to-real RL for a 3D-printed biped, with Isaac Sim / Isaac Lab training and a Raspberry Pi runtime
- Foosball_CU: a robot that plays foosball, with RL agents trained in simulation (NeurIPS 2024)
- LaunchLens: a synthetic market simulator built on LLM-driven consumer agents
- Smart_Movie_Setup: ambient LED lighting synced to on-screen video