Dinesh Thakur

Sr. Staff Robotics Engineer @ Burro

About

As a full-stack roboticist, I am driven by the challenge of transforming complex theoretical concepts into robust, real-world autonomous systems that work at scale. My career has been a journey from foundational robotics research at the University of Pennsylvania to leading high-stakes commercial projects in the fast-paced startup world. This unique blend of academic rigor and agile industry execution allows me to architect solutions that are both innovative and practical. My expertise lies in architecting and optimizing the entire robotics stack, from low-level driver enhancements to high-level strategic roadmapping. I am fluent in C++, Python, and the ROS/ROS2 ecosystems. With over a decade of hands-on experience in perception, planning, and control systems for ground and aerial robots, I thrive on solving the toughest challenges in autonomy. I have a proven track record of leading cross-functional teams, mentoring engineers, and setting the technical vision for complex projects. Feel free to connect with me to discuss the future of robotics, strategic development, or potential collaboration opportunities.

Country

United States

City

Greater Philadelphia

Industry

Information Technology & Services

Skill

Sensor Integration, Teamwork and Collaboration, Project Management, Swarm Intelligence, Micro Aerial Vehicles (MAV), ZigBee, Visual Inertial Odometry, Graph Neural Networks, Docker, Sensor Fusion, ROS2 (Galactic, Humble) , ARM Cortex-M, Robotics, ROS, C++, Python, Matlab, Computer Vision, Research, Embedded Systems

Experience

Sr. Staff Robotics Engineer

Burro

2025-4 - Present · 1 yr 6 mos

Staff Robotics Software Engineer

Burro

2024-4 - 2025-4 · 1 yr 1 mo

Consulting Director - Robotics

LF Intelligence Limited

2022-10 - 2024-4 · 1 yr 7 mos

Philadelphia, Pennsylvania, United States

I spearheaded and directed technical efforts to optimize autonomous lawncare robot performance, resulting in deployment improvements across 1200 units. Key achievements included enhancing sensor fusion pipelines for improved reliability and streamlining CI/CD processes using Docker for faster deployments. These improvements include: CPU Optimization: Led CPU optimization for Novabot autonomous lawncare robots (Horizon Robotics X3M AI SOC, quad-core 1.2 GHz Cortex A53), enabling concurrent operation of all critical robotic ROS2 nodes on the X3 M platform (previously >600% CPU). Achieved substantial per-core savings across components: ‣ 45% CPU reduction per camera (two cameras): Offloaded JPEG compression to hardware video codec. ‣ 45% total CPU saving on subscriber side (15% per node for QR code detection, VIO, perception): Implemented CycloneDDS shared memory and zero-copy communication. ‣ 10% CPU saving per node (two nodes): Implemented Image Pyramids using X3M hardware acceleration. ‣ 30-35% CPU saving: Offloaded Geometric Distortion Correction (GDC) to hardware IPU. ‣ 25% CPU saving: Improved TOF camera driver performance on Arm A53 through collaboration with PMD Technologies. Visual Intertial Odometry (VIO) Solution Improvements: ‣ Ensured accurate synchronization of IMU, camera images, and GPS data by implementing precise timestamping mechanisms, critical for robust and drift-free VIO, enabling precise robot localization and navigation. ‣ Reduced latencies in sensor data acquisition by strategically adjusting thread priorities, leading to more timely data fusion within the VIO pipeline and resulting in faster, more accurate pose estimation, especially during dynamic robot movements. ‣ Significantly decreased timestamp jitter in sensor data from 16ms to an average of 0.5ms that resulted in smoother and more reliable VIO estimates, crucial for accurate control and path following of the robot's chassis.