Sourabh Tiwari

About

Robotics software engineer from Bigbot who led localization and perception for an outdoor autonomous delivery robot, building ROS2/Nav2-based autonomy with EKF sensor fusion (IMU/odometry/GPS) and perception-driven dynamic costmaps. Experienced taking systems from Gazebo simulation to real-robot deployment, optimizing real-time behavior via logging-driven debugging and latency reduction, and integrating heterogeneous comms (MAVROS/MAVLink, UART/CAN, MQTT) for distributed and multi-robot setups.

Experience

Education

Key Strengths

• Built and integrated core autonomy stack components (localization, navigation, perception) for outdoor delivery robot
• Deep ROS2/Nav2 implementation experience (custom nodes, messages, costmap layers)
• Robust outdoor localization via EKF fusion of IMU, wheel odometry, and GPS
• Improved obstacle avoidance using continuously-updated dynamic costmaps driven by perception
• Real-world debugging approach: log/inspect sensor + costmap data, identify false obstacles/localization drift, tune filters
• Reduced perception latency to maintain consistent real-time updates
• Led localization and perception team and drove integration with broader software stack
• Designed reliable heterogeneous robot communications across edge computers/microcontrollers using ROS2 + MAVLink/UART/CAN/MQTT
• Strong sim-to-real workflow using Gazebo plus Dockerized deployments and automated build/test validation
• Built a custom ROS2 perception stack for autonomous sidewalk delivery robot using depth camera + 2D LiDAR (no 3D LiDAR)
• Designed modular ROS2 architecture with clean node interfaces and performance-optimized custom message types
• Reduced perception latency from ~90ms to ~30ms via vision/depth fusion into a single occupancy grid
• Integrated YOLO segmentation into a real-time ROS2 pipeline on NVIDIA Jetson using TensorRT and shared-memory transport (25–30 FPS)
• Improved navigation reliability: reduced navigation failures by 70%+; system adopted into next prototype
• Robust localization via EKF tuning to handle inconsistent sensor rates and noise (GPS drift, IMU bias) with ICP+GPS+IMU fusion
• Production-minded reliability practices: profiling early, time-bounded modular nodes, watchdog/heartbeat monitoring, real-world stress testing

Reference Highlights

• Strong technical skills
• Strong programming skills
• Strong ROS/ROS2 experience
• Able to build robotics software from scratch
• Good debugging skills
• Works collaboratively with the team
• Effective cross-functional collaborator (hardware/firmware)
• Interpersonal skills
• Leadership quality
• Hardworking
• Best fit for robotics software engineer role
• Expert in robotics (especially AI)
• Designed the full ROS framework for the robots
• Solved AI challenges easily
• Strong debugging ability
• Hands-on testing mindset (tested robots many times)
• Effective at real-time navigation and obstacle avoidance problems
• Able to coordinate/manage a multi-robot platform (around 10 robots)
• Strong teamwork and cross-disciplinary collaboration
• Values accuracy/precision when working with hardware
• Creative thinker (thinks differently)
• Cost-effective/efficient solutions (good results with cheaper sensors)
• Highly talented
• Would be rehired