Senior Machine Learning Perception Engineer

Mytra; City of Brisbane; On-site; AUD 140,000 - 220,000; Full time

Job summary

Mytra is seeking a Senior ML Perception Engineer to lead perception across our distributed robot fleet. You will build perception models for object detection, depth estimation, and scene understanding, while owning data pipelines and deployment.

Collaboration with camera, robotics, and safety teams will shape the perception roadmap and ensure robust on-bot performance. The role spans data collection to deployment, including sim-to-real, calibration, and optimization for real-world robots.

Qualifications

• BS in CS or ML with hands-on experience in computer vision or ML perception systems.
• Strong Python and PyTorch skills; comfortable with C/C++ for on-device or performance-critical code.
• Experience building or operating vision data pipelines and labeling workflows.

Responsibilities

• Develop perception models for fleet, including object detection and depth estimation.
• Own data ingestion, labeling workflows, and dataset curation for training and validation.
• Optimize models for on-bot compute and integrate perception outputs across the robotics stack.
• Collaborate with camera, robotics, and safety teams to shape the perception roadmap.

Skills

Python

PyTorch

C/C++

Data pipelines

Camera calibration

Education

Bachelor's degree in CS/ML

Tools

ROS

TensorRT

Simulation tools

Job description

• As a Senior ML Perception Engineer at Mytra, you'll be a key member of our Computer Vision team, building the perception stack that gives our distributed robot fleet its situational awareness.
• This is a hands-on role that spans the full problem: the models that let our robots detect pallet trays and other bots and estimate distance to them, the simulation and data pipelines that produce and validate those models, and the integration that gets perception onto the robot and into the hands of the motion and safety systems
• Perception feeds everything downstream: collision avoidance, localization, and safety all depend on how well our robots see and interpret their environment, so this role has real impact across the robotics stack.
• You’ll own significant perception workstreams end to end, from data collection through deployment, and work closely with the camera, robotics, and safety teams
• Build and improve perception models for the fleet, including object detection (pallet trays and bots), depth estimation (using both learned and geometric methods such as projection geometry and calibration), and scene understanding that power collision avoidance and other on-bot behaviors
• Own the vision data engine end to end: ETL pipelines that ingest on-bot and rollbot logs, route data through cloud storage and labeling workflows, and produce curated, training-ready datasets on demand, so improvements are measurable and reproducible
• Advance simulation and sim-to-real: train and evaluate models in simulation, validate them in software-in-the-loop, and systematically close the gap to real robot data
• Develop camera and time-of-flight based localization that feeds the safety system, improving the robustness of fleet localization
• Deploy and integrate perception on the robot: optimize models for on-bot compute, publish outputs to downstream consumers, implement consumer handling of perception inputs and build the automated processes to release and monitor models across the fleet
• Contribute technical leadership: author design docs, review code, mentor engineers, and help shape the perception roadmap in partnership with the camera/sensing, robotics, and safety teams

5+ years of directly relevant experience building and shipping computer vision or ML perception systems (BS required; MS or PhD preferred)The ability to work across disciplinary boundaries, collaborating with software, controls, and safety engineers to debug real robotic systemsHands-on experience with cameras and depth/ToF sensors, including data collection and calibrationComfort with ambiguity and an early-stage mindset: you're excited to shape an immature, high-impact area, and you're willing to realize mistakes and pivotStrong Python and PyTorch skills, and comfort with C/C++ for on-device or performance-critical codeExperience building or operating vision data pipelines and labeling workflows: dataset creation, curation, and assisted or auto-labelingA track record of shipping perception to production on real systems, with the data-centric debugging instincts to know why models break and how to fix themSimulation and sim-to-real experience (Isaac Sim, Webots, Gym, MuJoCo, or similar)Strong applied computer vision and deep learning skills (detection, segmentation, tracking, depth estimation) together with solid classical and geometric CV, including camera calibration and projection geometryEdge and embedded model optimization (quantization, pruning, distillation, TensorRT or similar) for resource-constrained on-robot computeFamiliarity with vision data and observability tooling (Foxglove, GCP, Encord, or equivalents)Foundation models, world models, or vision-language approaches applied to perception or simulationRobotics and systems integration experience (ROS or similar, CAN, pub/sub frameworks) and deploying perception across distributed robot fleets