Computer Vision Engineer

In this role, you will blend computer vision, AI, sensor fusion, and embedded engineering to deliver robust, field-validated autonomous capabilities. You will shape key architectural decisions and help bring cutting-edge autonomy features into real-world deployment.

Key Responsibilities

• Design, implement, and maintain core autonomy modules that integrate sensing, perception, state estimation, mapping, and planner interfaces into a cohesive real-time system.

• Develop high-performance computer vision pipelines (classical + AI-based) for detection, segmentation, tracking, and scene understanding, ensuring reliable operation on embedded hardware.

• Build multimodal perception systems that fuse camera, LiDAR, radar, and IMU data into accurate, navigation-ready environment representations.

• Deploy, optimize, and maintain autonomy software on embedded platforms (Jetson AGX/Orin), including TensorRT optimisation, cross-compilation, CUDA acceleration, and performance tuning for real-time execution.

• Own sensor bring-up, configuration, calibration, and synchronization (camera, LiDAR, radar, IMU, GPS), ensuring accurate and stable data for downstream modules.

• Ensure system-level robustness and safety by maintaining strict latency budgets, deterministic behaviour, numerical stability, and fall-back mechanisms for degraded sensing conditions.

• Conduct field trials, capture datasets, analyse system performance, and drive iterative improvements across sensing, perception, fusion, and planning layers.

• Debug deep autonomy stack issues including timing mismatches, calibration drift, concurrency conflicts, synchronization faults, and hardware–software integration challenges.

• Build deployment-ready autonomy systems using ROS/ROS2, Docker, systemd services, and reproducible build pipelines tailored for embedded platforms.

• Collaborate with mechanical, electronics, and systems teams to align autonomy software capabilities with real-world hardware constraints and vehicle dynamics.
• Contribute to autonomy architecture evolution, influencing design decisions, modularisation strategy, safety mechanisms, and long-term capability roadmap.

Required Technical Skills

• Required Bachelor's degree in Robotics, Computer Science, Mechatronics or related field.

• Master's or PhD in Robotics, Autonomous Systems, AI/ML, Computer Vision, or Control Systems is preferred.

• Strong proficiency in modern C++ (14/17/20) and Python for building production-grade robotics, CV, and autonomy software.

• Deep understanding of computer vision fundamentals (feature-based vision, geometric methods, multi-view geometry) and AI-based perception using PyTorch.

• Practical experience deploying and optimising perception models on embedded GPU platforms (Jetson Xavier/Orin or similar).

• Hands-on expertise with Triton, TensorRT, mixed-precision inference, Numba-JIT, CUDA kernels, and real-time optimisation techniques.

• Strong command of ROS/ROS2, TF transforms, message passing, node graph architecture, and middleware integration patterns.

• Extensive experience with robotics sensor integration including RGB/stereo/depth cameras, LiDAR, radar, IMUs, and GPS—covering calibration (intrinsic/extrinsic), synchronization, timestamps, and data integrity.

• Knowledge of core autonomy concepts: mapping, costmap generation, scene representation, obstacle detection, and planner interfacing.

• Solid grounding in Linux systems, multithreading, memory optimisation, real-time constraints, and system-level debugging workflows.

• Experience with Docker, cross-compilation toolchains, embedded deployment pipelines, and CI/CD systems for robotics software.

• Familiarity with simulation tools (Gazebo, CARLA, Isaac Sim) for developing reproducible test setups and automated validation.

• Ability to troubleshoot complex issues across perception, fusion, hardware interfaces, timing, concurrency, and algorithmic edge cases.

• Strong understanding of coordinate frames, transforms, camera models, rigid-body geometry, and numerical optimisation methods.

• Experience using logging frameworks, telemetry tools, performance profilers, and methods for long-duration stability testing.