Help us build a system that decides when to feed – on its own

OptoScale’s underwater cameras record fish in real time, around the clock. Our machine learning models turn the images into measurements of weight, lice, health and behaviour, and show how the fish respond when the feed comes in. Now we want to close the loop.

Feeding thousands of fish is a control problem with almost everything that makes control hard. You only see part of what’s going on. The feedback is noisy and delayed. Currents, temperature and weather keep shifting the conditions. And at the other end of the loop are living fish, whose appetite, stress and behaviour change from one day to the next.

We already have the sensors, the data and the models, and the first parts of the decision layer are in place. The next step is to make it autonomous, and we’re looking for someone to help us get there.

You’ll join our AI & Optics team in Trondheim, a group with backgrounds in physics, mathematics, cybernetics and marine technology.

What you’ll do

• Shape and develop the control strategy for autonomous feeding: how sensor data turns into feeding decisions

• Estimate the state of the fish and their environment from noisy, delayed data, and forecast how both develop over time

• Work with our machine learning team to turn model outputs into reliable inputs for feeding decisions

• Write Python that runs in production, not just in notebooks

• Follow your work from first idea to deployment, and keep improving it once it’s running

Who you are

You have an MSc or PhD with a strong foundation in control theory, cybernetics or autonomous systems. You write strong Python, and you’re fluent in English, written and spoken.

You take ownership and finish what you start, also when it gets hard. You know when to pull others in and when to get on with it yourself. And you’d rather see something work in the sea than look good in a simulation.

What gives you an edge

• At least three years in research or development (a PhD counts)

• Statistical modelling, data science or forecasting

• Reinforcement or imitation learning

It’s also a plus if you’ve built something that already runs in a real product, worked with transformers, generative models, computer vision or deep learning, or know aquaculture.

About us

We operate across Norway, Sweden and Chile. We build our own hardware and software, from the optics and electronics in the cameras to the algorithms and the portal our customers log into. So you’re not working around a black box. If your controller needs better data, you can sit down with the people who build the sensors.

We’re small enough that what you build gets used, and specialised enough that someone will push back on your thinking.

Recruitment process

We are committed to fair and inclusive recruitment. To reduce bias, applications are first reviewed anonymously. Selected candidates will be invited to take an aptitude test and a personality test (Matrigma and MAP from Assessio), and we may carry out background checks as part of the process.