Principal Robotics AI Engineer (Perception & Autonomy Systems)

Job Overview

AI.DA Strategic Technology Centre (STC)’s Next-gen Edge AI & Robotics (NEAR) Lab operates a heterogeneous fleet of legged robots, drones, mobile manipulators and other robot types for autonomous operations in unstructured, GPS-denied environments across different applications. We are looking for a Principal Engineer to own the perception layer across this fleet: real-time 3D scene understanding, multi-sensor fusion, and the systems engineering that makes perception work reliably across platforms with different sensor payloads and compute budgets.

This is an engineering-first role with active research participation. You ship real-time perception systems that survive field conditions across multiple platform types. You will work with academic collaborators to translate research into actual applications. The lab has strong academic partnerships forming in cooperative perception and semantic mapping; you will be the technical counterpart on our side of those collaborations.

You will partner closely with the lab's existing mapping and SLAM lead, who guides LiDAR-inertial odometry and map-building. Your focus is perception proper: what the robot sees and understands about its environment, and the systems architecture that makes it work across the fleet.

This role reports to the Lab Director within the Capabilities pillar.

Perception & Scene Understanding

• 3D Scene Understanding: Build and ship the perception pipeline across the lab's platforms: object detection, instance segmentation, terrain classification, traversability estimation. Real-time C++ on edge compute, with clean interfaces to downstream mapping, planning, and control.
• Terrain Perception and Traversability: Own the terrain understanding layer: elevation mapping, traversability scoring, and the integration of classical geometric methods with learned approaches (foundation-model features, self-supervised traversability). Different platforms have different mobility envelopes; the traversability layer must account for this.
• Cooperative Perception: Drive the lab's multi-robot cooperative perception work: shared scene representations, distributed inference, bandwidth-aware feature sharing across heterogeneous platforms. This is a core research direction with our academic partners.
• Multi-Sensor Fusion and Calibration: Integrate perception across different sensor configurations (RGB-D, thermal, fisheye, LiDAR and IMU) and different platforms. Produce equivalent, reliable scene representations regardless of the sensor payload. Own the multi-sensor calibration pipeline: intrinsics, extrinsics management across platform changes, and time synchronisation.

Wider Scope

• Programme Delivery Support: Provide C++ systems depth for real-time autonomy stacks across programme delivery, particularly where perception, sensor fusion, or edge compute performance are the bottleneck.
• Academic Partnership: Serve as the lab's technical counterpart for cooperative perception research collaborations.

Required Qualifications

• Education & Experience: PhD strongly preferred in Robotics, Computer Science, Computer Vision, Electrical Engineering, or related field. 8+ years of post-graduate experience in perception for mobile robotics, autonomous vehicles, or field robotics. Master's-level candidates with strong industry-research track records are welcome too.
• 3D Perception Depth: Hands-on experience with 3D object detection, semantic segmentation, or scene understanding on real robots, not just benchmarks. You've dealt with sensor degradation, domain shift, and real-world failure modes.
• Real-Time C++ Systems: Strong C++17/20 proficiency with demonstrated experience building real-time perception pipelines on edge compute. This is a core requirement.
• Multi-Platform Perception: Direct experience shipping perception systems across more than one robot platform type, integrating different sensor payloads with robust handling of varying FOV, timing, and degradation.
• ROS 2: Strong working knowledge of ROS 2 including frame conventions, multi-robot architectures, and distributed systems.
• Real-Robot Track Record: Multi-year pattern of deploying perception systems onto real hardware under field conditions.

Preferred Qualifications

• Multi-Robot Field Challenge Experience: Direct involvement in multi-robot field challenges like MBZIRC.
• Cooperative / Multi-Robot Perception: Prior work on distributed perception, multi-robot scene fusion, or decentralised SLAM backends.
• Neural Scene Representations: Experience with 3D Gaussian Splatting, NeRF, or neural implicit representations for dense scene reconstruction, mapping, or navigation. Awareness of how these representations serve as coordination substrates in multi-robot systems.
• Elevation Mapping and Traversability: Experience with 2.5D elevation mapping, GPU-accelerated terrain analysis, or learned traversability estimation for mobile robots.
• LiDAR-Inertial SLAM Depth: Working depth in FAST-LIO2, GLIM, ORB-SLAM3 or similar, beyond integration, into development and adaptation.
• Scene Graphs and Semantic Representations: Experience with hierarchical scene representations for multi-robot coordination.
• Foundation Models for Perception: Experience deploying open-vocabulary detection or VLMs on real robots with awareness of latency and reliability constraints.
• Calibration Discipline: Experience with multi-sensor calibration, extrinsics management across platform changes, and calibration-as-code.
• Research Output: Peer-reviewed publication at ICRA, IROS, CoRL, CVPR, or ECCV.
• Field Robotics: Experience in GPS-denied, degraded-visual, or outdoor unstructured environments.
• AI-Assisted Engineering Judgement: Fluent with modern AI coding assistants, with clear judgement on when to use them and when to read code directly.