Experience
My professional journey in robotics and research.
Relativity Space
Perception Researcher
June 2025 - Present
Long Beach, CA
- Built a motion-capture system from scratch, leveraging traditional Structure-from-Motion techniques.
- Achieved ~10 micron global accuracy in rigid-body localization, outperforming OptiTrack’s published 150 micron benchmark by an order of magnitude (IR Camera based).
- Developed a fully autonomous, self-calibrating pipeline using a custom calibration stand (no human in the loop) for faster and more robust initialization, with self-healing extrinsics based on epipolar geometry and probabilistic reprojection-error estimation.
Caltech (AMBER Lab)
Robotics Researcher
November 2025 - Present
Pasadena, CA
- Conducted reinforcement learning research for humanoid locomotion in Isaac Lab, developing LIP-inspired policies for velocity tracking and foot-contact regularization.
- Designed reward constructions emphasizing phase-consistent gait generation, center-of-mass stability, and energy-efficient motion under disturbances.
- Implemented procedural maze terrain generation in Isaac Lab using DFS-based layout generation and Wave Function Collapse (WFC) for edge-consistent ramp, stair, and flat mesh assembly.
.406 Ventures
Fellow
May 2025 - Present
Boston, MA
- Selected as 1 of 15 Student Founders Nationwide Composing Class XVII.
- Builder-first venture fellowship.
Air Force Research Laboratory
Computer Vision Researcher
June 2024 - September 2024
Dayton, OH
- Automated digital twins manufacturing using OptiTrack motion capture and 3D scanning for robo-simulations.
- Integrated ROS2 into the digital twin pipeline and developed a custom Dockerfile for optimized deployment and consistent environments using Kernel-Based Virtual Machines.
- Created a 3D similarity score program via Iterative Closest Point and Hausdorff distance to validate YCB dataset.
- Fine-tuned Meta SAM2 with PyTorch and YOLOv8 to create custom weights, enabling Boston Dynamics Spot to identify lab-specific objects with 92.8% accuracy.
Caltech (Wierman Group)
Machine Learning Researcher
July 2023 - September 2023
Pasadena, CA
- Evaluated the impact of adversarial noise on CNN-based machine perception for autonomous driving, identifying a 40% accuracy reduction in critical image regions using Python and TensorFlow.
- Developed a CNN with 52,673 trainable parameters across four layers (convolution, max pooling, flatten, and dense) to assess noise effects using a dataset of 2,000 images.
- Enhanced the robustness of machine perception by retraining the CNN on noise-affected data, achieving a 20% improvement in accuracy for vehicle position identification under adversarial conditions.