Perception Engineer

GrazeMate, Inc.; Sydney; On-site; AUD 150,000 - 190,000; Full time

Benefits offered by this job

Competitive salary

Meaningful equity

Job summary

Brumby is building autonomous drones to herd cattle and monitor pasture. We seek a Perception Engineer to own how our drones see the world, train models, and deploy them on Jetson for cattle detection, counting and tracking in real time.

You will fuse camera, GNSS/IMU data and run edge AI pipelines while collaborating with autonomy engineers to translate perception into flight behavior, with regular field deployments from Sydney.

Qualifications

• Trained perception models deployed on Jetson with TensorRT/INT8.
• Built DeepStream or GStreamer pipelines in production.
• Strong sensor fusion across camera and GNSS/IMU with pose estimation.
• ROS 2 in production on real hardware (Jazzy, MAVROS).
• Experience with real field data and YOLO in limited compute.
• Experience with SITL, Gazebo, rosbag for testing.
• Degree in CS, robotics, mechatronics or EE, or related research.

Responsibilities

• Onboard perception that detects, counts, and tracks cattle from the air.
• Maintain edge inference pipelines on Jetson with real-time latency.
• Fuse camera, thermal, and GNSS/IMU data to geolocate animals.
• Bridge perception and autonomy so it informs flight behavior.
• Create simulation/replay tooling to validate before field tests.

Skills

Perception models

Jetson deployment

TensorRT INT8

DeepStream

Sensor fusion

ROS 2

YOLO

SITL Gazebo

rosbag

Field data

Education

CS/Robotics/Mechatronics/EE degree

Research background in perception/autonomy

Tools

DeepStream

GStreamer

ROS 2

MAVROS

Job description

The Company

Founded in 2025 by Sam Rogers, who grew up managing 6,000 head on a North Queensland cattle station, Brumby builds autonomous drones that move and monitor cattle across the world’s largest properties. We’re replacing helicopters and motorbikes with drones that herd cattle, count herds, and monitor pasture on their own. Helicopter mustering is expensive, dangerous, and still misses cattle. Our drones launch from a dock, fly the paddock, do the work, and come home.

We went through YC (W26) and raised a seed round from the best investors in the valley. Our team is small and highly technical, and we’re already working with some of the largest ranches in the world. Engineering is based in Sydney, but the work happens everywhere. Expect to write code at a desk, in a truck, and in the dirt. We work hard because our ranchers do.

We have a team of 6 which is growing steadily, including Computer Vision, Robotics, AI Engineers and a Chief of Staff. We are looking to add a Perception Engineer who will own how our drones see the world and act on it. This is a hands‑on role with real ownership, working closely with the founder and the rest of the engineering team to get our aircraft finding and mustering cattle on their own.

The Role

You will build the perception systems that let Brumby’s drones find, count, and track cattle from the air. That means training models that work in challenging real‑world conditions and using GNSS to georeference detections so every animal is correctly located. Beyond cattle, you’ll help the drone read the land itself, from checking fence lines and water troughs to understanding pasture conditions across a property.

All of it runs on the aircraft, on edge hardware with limited compute. You will take models from messy, real‑world flight footage through to fast, reliable inference on a Jetson. You’ll also be out on stations seeing how those models hold up in the field, then bringing that data back to make them better.

You will work closely with the rest of the engineering team, including those working on autonomy, so that what the drone sees translates cleanly into how it flies. Perception and autonomy sit close together here, with plenty of room to grow across both.

What we’re looking for

Must-haves

• You have trained perception models and deployed them on Jetson, using TensorRT and INT8 to hit a real latency budget

• You’ve built DeepStream or GStreamer pipelines that run in production

• Strong sensor fusion skills across camera and GNSS/IMU, including pose estimation, geo‑registration, and multi‑frame tracking

• You’ve run ROS 2 in production on real hardware, including Jazzy, DDS tuning, and MAVROS

• You’ve worked with real field data, including messy footage and bad labels, and trained models like YOLO in unusual domains where compute is limited

• Experience with simulation and replay tools like SITL, Gazebo, and rosbag to test and validate systems before they fly.

• A degree in computer science, robotics, mechatronics, or electrical engineering, or a research background in perception or autonomy

Strong advantages

• You’ve fused thermal or multispectral imagery with optical data

• You’ve worked on SfM, photogrammetry, or biomass estimation

• You understand how ArduPilot or PX4 behaves in AUTO and Guided modes, including its failure modes and edge cases.

• You’ve worked on path planning and area coverage, whether for survey, mustering, or exploring unknown terrain

• You’ve worked on drones or been part of a team building them

The Work

• Onboard perception that detects, counts, and tracks cattle from the air, day and night

• Edge inference pipelines on Jetson that run in real time with no connectivity

• Fusing camera, thermal, and GNSS/IMU data to geo‑locate every animal in the paddock

• The interface between perception and autonomy, turning what the drone sees into how it flies

• Simulation and replay tooling so we can test the full loop before it goes into the field

The Setup

• Based in our Sydney office five days a week

• Regular field deployments to cattle stations (domestic and potentially international)

• Full Australian working rights required

• Work directly with (and report to) the founder

• Competitive salary, meaningful early stage equity

Application deadline: October 30, 2026, 5PM AEDT.