Rituraj Navindgikar

About

Robotics software engineer with hands-on experience building Visual-Inertial SLAM and ROS2 sensor-fusion pipelines for autonomous warehouse forklifts (ArcBest), including rigorous calibration (AprilTags, Allan variance, temporal sync) and recovery features like pose injection. Also implemented RL-based local planning at RollNDrive using Isaac Sim with domain randomization to bridge sim-to-real, improving real-world navigation success back to ~90% after initial deployment.

Experience

Education

Key Strengths

• Built Visual-Inertial SLAM for autonomous forklift localization in challenging warehouse environments
• Deep calibration expertise (AprilTags, IMU Allan variance, extrinsics + temporal sync) with quantitative validation via reprojection error thresholds
• Robustness mindset: loop-closure strategy to mitigate drift; accounts for vibration/temperature/mechanical stress over time
• Applied RL for real-time local planning; trained ~10,000 episodes and handled moving obstacles (~0.5 m/s)
• Strong sim-to-real execution using domain randomization and conservative rollout; improved real-world success from ~80% back to ~90%
• ROS2 production integration skills: time synchronization tuning, timeout handling, degraded-mode operation with higher uncertainty
• Designed recovery mechanism via pose injection into localization/EKF update step for delocalization events
• Practical debugging/observability: rosbags, diagnostics, Foxglove visualization, and custom Python dashboards