Karan Shah

About

Robotics software engineer focused on production-grade autonomy in GPS-denied environments, building full navigation stacks (perception, EKF/UKF sensor fusion, planning, control) in ROS2. Integrated YOLOv8/semantic segmentation/RL policies into real-time NAV2 pipelines via a custom perception-aware costmap layer, with emphasis on deterministic control loops, embedded GPU performance, and robust system observability/fault tolerance.

Experience

Education

Key Strengths

• Built end-to-end autonomous navigation stack (perception, fusion, planning, control) for multi-terrain mobility
• Deep expertise in GPS-denied localization using EKF/UKF multi-sensor fusion
• Strong ROS2/NAV2 engineering: developed custom perception-aware costmap layer integrating ML outputs into planning
• Real-time robotics optimization: deterministic control/planning with asynchronous ML inference on embedded GPUs
• Systems-level thinking: closed-loop integration between localization, planner, and controller
• Practical handling of real-time latency spikes in perception pipelines impacting navigation stability
• ROS 2 + Pixhawk integration via MAVLink for UAV control
• Real-time sensor synchronization and latency/communication tuning
• EKF-based sensor fusion for navigation using depth cameras + IMU
• Debugging and optimizing obstacle avoidance using data-flow analysis and time-sync improvements
• Simulation-to-deployment workflow using Gazebo with ROS 2
• Containerizing and deploying ROS 2 systems with Docker and Kubernetes for scalable distributed services
• Coordinating heterogeneous robots (UAV + ground robot) with ROS 2 messaging and MAVLink command links