Founding Engineer, Computer Vision & Robotics

Mission

We want to automate human work in factories. We are starting with quality inspection.

Trenon is building a perception system that understands physical work as it happens.

Our first product uses first-person cameras worn by production workers. It observes manual assembly, identifies which parts are present, which process step is running, and whether the expected sequence is being followed. When something goes wrong, it flags the deviation before it becomes rework, scrap, or a defect that reaches the customer.

Quality inspection is the starting point. Over time, every deployment can also capture first-person data on how skilled workers perform physical tasks. Our longer-term hypothesis is that this becomes training data for robotic manipulation.

Understand human work. Verify it. Learn from it. Automate it.

We are two founders, backed by Entrepreneurs First, and are starting our first factory deployments in the US. You would be our first engineer.

What you’ll own

You will own the path from experimental perception to a production system that works reliably in factories.

You will

• Build real-world perception across detection, classification, tracking, temporal understanding, and 3D vision.
• Solve first-person perception under camera motion, changing viewpoints, occlusion, motion blur, and changing lighting.
• Own the synthetic-to-real loop with the CTO, turning real-world failures into better models, data, and evaluation.
• Build geometric vision across calibration, localization, pose estimation, and coordinate transformations.
• Train and debug PyTorch vision models across synthetic and real data.
• Build low-latency inference and deploy on NVIDIA Jetson and TensorRT.
• Deploy directly on factory floors and turn each deployment into reusable perception infrastructure.

Over time, the stack can extend toward 3D human and object trajectories, robot learning, and robotic deployment.

Who we’re looking for

You are unusually strong in computer vision, robotics perception, or applied ML and have made perception work outside a clean notebook or lab.

Strong Python, PyTorch, OpenCV, geometric vision, and real-time camera systems matter.

Experience with synthetic data and Sim2Real, SLAM or VIO, depth and 3D reconstruction, CUDA or TensorRT, ROS2, robot manipulation, or robot learning is a strong plus.

We do not care about years of experience, degrees, or employer names.

We care whether you can own an underspecified physical-world problem end to end and distinguish a model metric from a product promise.

Why this role is unusual

You are not joining an existing engineering team. You are helping create one.

There is no product manager between you and the factory, and no mature perception stack for you to inherit. You will work directly with both founders, deploy what you build in real plants, and have significant influence over the technical direction of the company.

The longer-term technical question is larger than quality inspection:

Can first-person perception of human work become the data and intelligence layer required to automate that work?

If that sounds like a problem you want to spend the next few years solving, we want to talk.