Engineered a comprehensive learning pipeline for complex robotic manipulation utilizing simulation design, reinforcement learning, and synthetic data generation. Formulated a novel hybrid RL method integrating suboptimal expert data, outperforming naive PPO and Imitation Learning baselines by a statistically significant 6%. Trained and deployed a visual servoing policy for a UR5 using Deep RL (PPO), achieving >50% success in unstructured physical environments. Improved target fruit detection accuracy by 27% (96% success rate) using SAM2 and a custom YOLO model. Deployed policies via ROS2 and MoveIt!, reducing experimentation time by 80% via Behavior Trees.

Developed scalable robotic simulation pipelines utilizing NVIDIA Isaac Sim and Isaac Lab for autonomous agricultural systems.

Developed network solutions on the 5G NR MAC team. Increased test coverage metric from 2.1 to 3.2 (exceeding required threshold of 3.0) using the Google Test framework. Earned the Software Professional Certification for proficiency in Data Structures and Algorithms.

Adapted state-of-the-art Image Inpainting models in PyTorch to Dynamic to Static LiDAR Reconstruction. Evaluated reconstruction quality by executing SLAM on reconstructed LiDAR output using ROS. Formulated the “LiDAR Quality Index” metric to rank LiDAR frame quality without reference ground truth.

Directed the student organization for the 2023-2024 academic year, coordinating events and advocating for graduate students.