Research

Engineering intelligent systems for robotics and autonomy.

Applied research in robot control, multi-agent coordination, and GPU-accelerated perception, designed around real constraints such as safety, latency, uncertainty, and sim-to-real transfer.

Advanced Robotic Technologies for Surgery (ARTS) Lab, Texas Robotics Aug 2025 - Present Austin, Texas

Robot Control and Motion Planning

Surgical Robotics Researcher

Safety-aware control: Developing collision-resistant, fault-tolerant, and singularity-aware control software for Franka Emika robots using Python and C++ ROS2 nodes.
Simulation validation: Building regression tests to evaluate motion-planning behavior before physical deployment.
Reusable robotics tooling: Published SE3Kit on PyPI, providing modular APIs for SE(3) transformations, rotations, and simulation workflows.
Depth stabilization for endoscopic perception: Developing camera-calibration and OpenCV pipelines in Python and ROS2, containerized with Docker for deployment on NVIDIA Jetson AGX Thor.
Spinal pedicle fixation planning: Building trajectory-planning workflows with Python and OpenGL using synthetic datasets and domain randomization for robust surgical guidance.

Texas Aerial Robotics Feb 2026 - Present Austin, Texas

Multi-Agent Autonomy

Swarm Robotics Engineer

35% lower end-to-end latency: Optimized ROS2 and RViz2 simulation workflows for faster perception-to-control execution.
Decentralized coordination: Engineering drone-swarm perception and control pipelines for reinforcement learning and multi-agent decision-making.
Performance benchmarking: Evaluating coordination strategies against latency, scalability, and simulation reliability constraints.

NOVA Self-Driving Research Lab Aug 2024 - May 2025 Richardson, Texas

Perception and Autonomous Systems

Autonomous Vehicle Engineer

Greater than 85% detection accuracy: Integrated YOLOv8 brake-light perception with CUDA-accelerated, cuDNN-optimized inference on Ubuntu Linux and Docker.
Real-time 3D perception: Processed distributed LiDAR telemetry through ROS2 and voxel-grid data structures for efficient point-cloud workflows.
Sim-to-real forecasting: Developed RNN-based fleet strategies and time-series experiments in CARLA for autonomous driving scenarios.
Dynamic occupancy-grid prediction: Built LSTM and Scene Transformer temporal models to forecast fused LiDAR and RADAR occupancy grids 5, 10, and 15 seconds ahead, deployed as modular Python ROS2 nodes and containerized with Docker.

YOLOv8 brake light detection annotation example 1 — Brake-light perception output

YOLOv8 brake light detection annotation example 2 — GPU-accelerated detection results

Dataset courtesy of IM Lab, Kookmin University via Roboflow Universe under CC BY 4.0. YOLOv8 annotations by Sai Peram. View dataset.