Computer Vision Engineer

We’re looking for someone who knows how to take computer vision beyond the experiment and make it work in the real world.

• You have built computer vision into a real product, with something you wrote running somewhere beyond your own laptop.

• Strong Python and PyTorch skills, backed by a solid understanding of classical computer vision. Segmentation, keypoints, pose estimation and basic point cloud work should all feel familiar, and you know when geometry is the right tool for the job.

• Experience deploying models on edge hardware, whether that is a Jetson board, a camera module with an onboard NPU, or a single-board computer with an accelerator. You understand the realities of latency, memory and thermal limits.

• Experience with depth sensing, whether stereo, time of flight or structured light. You are comfortable with calibration and understand the challenges that come with noisy depth data.

• A pragmatic approach to data. You understand that a well built ground truth set is worth more than a clever architecture, and you have used experiment tracking and some form of dataset versioning so you know which model saw which data.

• Solid engineering habits. Version control, containers, continuous integration and enough testing that someone else can pick up your code and get it running.

• Real fluency with generative AI, in both directions. You use agentic coding tools to move faster and understand where vision foundation models genuinely add value — from reducing annotation effort to detection and pretrained depth models. We value engineers who can get strong results from a few thousand labelled animals rather than waiting for a hundred thousand.

Experience with depth SDKs such as Luxonis DepthAI, Stereolabs ZED or Intel RealSense, with AWS, Databricks or MLflow, with C++ on latency sensitive paths, or with livestock and agriculture, would all be highly regarded.