PhD researcher at IHMC in Pensacola, FL, working at the intersection of classical control theory and deep reinforcement learning for humanoid robot locomotion and recovery. My broader interests span mathematical optimization, physics-informed learning, and the principled design of autonomous systems.
Education
| Institution | ||
|---|---|---|
| Ph.D., Robotics | IHMC / University of West Florida, Pensacola FL | 2024 – present |
| M.Sc., Applied Mathematics | Northeastern University, Boston MA | 2019 – 2021 |
| B.Tech., Mechanical Engineering | NMIMS, Mumbai India | 2015 – 2019 |
Research
I am a PhD researcher at the Institute for Human and Machine Cognition (IHMC) in Pensacola, FL, one of the world’s leading humanoid robotics laboratories.
My doctoral work focuses on physics-embedded reinforcement learning for humanoid balance and recovery. Decades of classical control research produced compact, interpretable representations of balance such as capture point, centroidal momentum, and wrench-feasibility regions, that modern RL has largely discarded. I embed these physical invariants directly into the training process as privileged critic inputs and reward structure, enabling humanoid robots to discover emergent multi-contact recovery behaviors without any prescribed contact schedule.
The target is a single policy governing the complete loco-manipulation recovery loop: stable walking, external disturbance, fall, multi-contact stand-up, and task resumption.
Experience
PhD Researcher, IHMC / University of West Florida (Pensacola, FL)
- Physics-embedded RL for humanoid robot balance and recovery
- Asymmetric actor-critic architecture with privileged balance signals: capture point, CoM, centroidal momentum
- 93.4% fall recovery rate with emergent multi-contact strategies across elbows, knees, and forearms
- Lab website
Research Engineer, DEKA Research & Development Corp. (Manchester, NH)
- Autonomous Delivery Robot: deep learning sensor fusion for global occupancy grids using lidar, radar, and cameras
- Insulin Pump Enhancement: reinforcement learning and transformer methods for adaptive insulin delivery
- Infusion Pump Analysis: vision-based liquid flow rate estimation
ML Research Assistant, Nano-medicine Center, Northeastern University (Boston, MA)
- Automated MRI classification for Alzheimer’s disease with 83% accuracy
- Segmentation models for gray matter characterization from structural MR images
Technical Skills
| Programming | Python, C++, Java, MATLAB |
| Simulation | Isaac Sim, IsaacLab, MuJoCo, ROS, sim-to-real transfer |
| Control Theory | classical controls, capture-point dynamics, centroidal momentum, wrench-feasibility analysis |
| Machine Learning | PyTorch, CUDA, reinforcement learning, transformer architectures, Scikit-Learn |
| Tools | NumPy, Pandas, OpenCV, Linux |
Robots
Some of the platforms I have worked with at IHMC and DEKA Research.
Music
I play alto saxophone with PBC Band, a community band based in Pensacola. Playing live with other musicians is one of the most enjoyable parts of my week.
Outside the Lab
Chess, running, hiking, photography, board games, puzzles. Happy to talk about any of these.
Selected Projects
RUL-Aircraft predicts the remaining useful life of aircraft engines using deep learning on the NASA C-MAPSS dataset.
YOLO v3 Object Detection reimplements the YOLO v3 algorithm from scratch in PyTorch, with custom data augmentation and evaluation on the MS COCO dataset.
Patterns in Capset is a combinatorics research project exploring attribute distributions in the cap set problem.