Resume
Ayushman Choudhuri
Robotics & Computer Vision Engineer
Profile
Robotics Software Engineer with 5+ years of experience building production robotic systems at scale. Specialized in low-latency 2D/3D object detection, multi-sensor fusion (camera, LiDAR, IMU), and embedded deployment on NVIDIA Jetson platforms. Skilled across the full stack, from camera-based ROS2/C++ perception pipelines and model quantization to MLOps for production edge deployment. M.Sc. in Robotic Systems Engineering, RWTH Aachen University.
Experience
- Own QualiCam, a ROS2 and NVIDIA Jetson–based real-time optical perception system for agricultural harvest quality assurance, end-to-end from architecture through field deployment.
- Own the end-to-end MLOps pipeline (datasets, auto-annotation, training, evaluation), cutting manual effort by 85% and accelerating model release cycles.
- Improve the ROS2 object detection pipeline and custom vision algorithms for the Farmsort.one optical sorter, enabling low-latency classification and sorting of agricultural produce.
- Lead deployment and maintenance of production ML models in the field, resolving customer issues to ensure stable, efficient system performance.
- Developed and field-validated a standalone harvest quality assurance system (QualiCam) at customer farms to assess crop quality in real-world conditions.
- Developed quantization and deployment pipelines for YOLO-based object detection models using TensorRT on NVIDIA Jetson Orin NX platform.
- Spearheaded development of a stereo vision–based 3D object detection pipeline for an autonomous harvester, enabling real-time collision avoidance. Showcased at Agritechnica 2023.
- Benchmarked and optimized state-of-the-art detection models on custom agricultural datasets, achieving 5× faster inference through ONNX and HailoRT quantization on Hailo8 edge processors.
Institut für Kraftfahrzeuge (ika), RWTH Aachen University · Aachen, Germany
- Developed an attention map–based explainability methodology for real-time, transformer-based 3D object detection, resulting in a peer-reviewed publication at IEEE ITSC 2026.
- Developed a ROS2-based pipeline to ingest LiDAR point clouds and display 3D object detection as well as saliency maps in real time.
RWTH Aachen University · Aachen, Germany
- Developed a perception pipeline for safe mobile robot operation in construction environments, focusing on stereo vision–based 3D object detection and LiDAR point cloud compression.
- Designed and deployed a closed-loop LiDAR tilt system to increase vertical field of view by 100% for close-range applications; ROS package written in C++ and deployed on NVIDIA Jetson Xavier.
Synedyne Systems · Bangalore, India
- Developed and implemented ML-based calibration algorithms on an edge device to estimate payload of a self-loading cement mixer truck.
- Achieved weight estimation accuracy of 98.5% with a maximum payload of 800 kg.
Indian Institute of Science · Bangalore, India
- Designed and developed control software (C/C++) and a robotic test bed for precise liquid dispensing for composite manufacturing at the Department of Aerospace Engineering.
- Achieved accuracy of ±10 microliters using a MEMS-based flow sensor.
Agilebot Automation · Bangalore, India
- Worked on AS/RS robotic system design for automating processes in medium and large scale warehouses.
- Developed chassis and frame design for a pick-and-place autonomous rover module for medium scale warehouses.
Education
RWTH Aachen University · Aachen, Germany
- Focus areas: computer vision, robot perception, autonomous systems
- Master's thesis: Explainable Transformer-based 3D Object Detection in LiDAR Point Clouds
- Conducted at Institut für Kraftfahrzeuge (ika) · Advisor: Till Beemelmans, M.Sc.
Manipal University · Manipal, India
- Active member of Team Robomanipal — ABU Robocon 2015 & 2016
- Co-founded Strange Matter Robotics during undergraduate studies
Research Publications
[1] T. Beemelmanns, S. Sharifi, M. Mehrotra, A. Choudhuri, L. Eckstein. Towards Trustworthy and Explainable AI for Perception Models: From Concept to Prototype Vehicle Deployment. IEEE International Conference on Intelligent Transportation Systems (ITSC), 2026. arXiv:2605.16087
Technical Stack
Programming
Frameworks & Libraries
MLOps
Developer Tools
Languages