CV

PhD Researcher and Research Assistant

University at Buffalo (SUNY), DRONES Lab

Researching safe planning, diffusion-based planning, learned dynamics, and embodied autonomy for robotic manipulators and autonomous systems.

• Developing collision-free trajectory generation for multi-DoF robot arms using Control Barrier Functions, Control Lyapunov Functions, arm kinematics, workspace constraints, and CLF-CBF-QP solvers in CasADi.
• Developing trajectory-level constraint-aware safe sampling methods that integrate closed-form CBF corrections into diffusion-model sampling updates for fast, safety-aware planning.
• Designing diffusion-based trajectory planners for hydraulic excavators using learned forward and inverse LSTM dynamics models over joint angles, velocities, pressures, and control inputs.
• Using learned inverse dynamics to convert MoveIt state trajectories into observation-control demonstrations and forward-dynamics guidance to promote physically coherent Gazebo and hardware execution.
• Serving as Field and Software Lead for EARTH, a three-year Moog-funded project developing autonomous excavators.
• Contributed Nav2 differential-drive navigation, MoveIt2 multi-link arm trajectory planning, Gazebo validation, optimized waypoint deployment on physical hardware, ROS2 lifecycle node management, and CAN bus interfacing.

Member Technical Staff, Software Development Engineer

Zoho Corporation

Built backend systems for cloud repository management and search.

• Led end-to-end development of Zoho's Cloud Repository Management System.
• Engineered asynchronous search and retrieval pipelines that reduced API response time by 60 percent.
• Integrated polyglot microservices with gRPC and Protobuf and redesigned database architecture to reduce redundancy by 70 percent.

Doctor of Philosophy

University at Buffalo (SUNY), DRONES Lab

Computer Science

Master of Science

University at Buffalo (SUNY)

Engineering Science: Robotics

Bachelor of Technology

SASTRA Deemed University

Mechanical Engineering

Excavation Autonomy with Resilient Traversability and Handling

ICRA Workshop on Field Robotics

Workshop paper on EARTH, a framework for autonomous excavators and earth-movers.

A Constraint Aware Framework For Safe Sampling

• Developing a deterministic safe sampling framework that integrates a closed-form trajectory-level CBF correction directly into diffusion-model sampling updates.
• Formulated a single smooth trajectory-level CBF by aggregating obstacle- and waypoint-level constraints through nested softmin approximations.
• Derived a first-order DPM-Solver discretization of the safety-augmented probability-flow ODE and delayed CBF corrections until final denoising steps to reduce local-trap formation.
• Demonstrated 100 percent trajectory safety across evaluated Maze2D settings and a 91 percent safe-success rate in dense PointMass2D while generating trajectories in 0.05-0.07 seconds.

Diffusion-based Trajectory Planning for Excavators with Learned Dynamics Models

• Designing a diffusion-based trajectory generation framework with learned dynamics models to generate dynamically coherent observation-control trajectories for hydraulic excavators.
• Engineered forward and inverse LSTM dynamics models over rolling histories of joint angles, joint velocities, hydraulic pressures, and control inputs for boom, arm, and bucket motion.
• Using learned inverse dynamics to convert MoveIt state trajectories into observation-control demonstrations and forward-dynamics guidance to promote physically coherent Gazebo and hardware execution.