Junior ML Data Engineer (Computer Vision)

Our models are only as good as the images they learn from. We are hiring a junior engineer to own the data behind our computer vision: collecting it from plants, cleaning it, labeling it and making sure every training set is one we can trust.

What we do

Wasteer puts computer vision and machine learning on the waste stream at European waste-to-energy plants. Cameras watch the bunker and the crane. Our models detect contaminants, classify the waste and guide how the bunker is mixed and fed. Paying customers run plants that handle 16 percent of Europe's waste, so the data you prepare ends up in production models running 24/7.

What you will do

• Collect image and video data from camera feeds at customer plants and keep it organised, versioned and findable.
• Clean datasets: remove duplicates, broken frames and near-identical images, and balance rare classes.
• Label images and video for object detection and segmentation, and write the labeling guidelines others follow.
• Run quality checks on labels, find disagreements and mistakes, and fix them.
• Use our models to pre-label data and find the images they get wrong, so labeling time goes where it matters.
• Build small Python tools and scripts that make all of this faster.
• Work closely with our ML engineers on what data the next model needs.

What we look for

Must have

• Python. You can write scripts that process thousands of files without babysitting them.
• Care for detail. A wrong label in a training set bothers you.
• Basic understanding of how a computer vision model learns from labeled data.
• A degree or projects in computer science, data science, engineering or a related field, or equivalent hands-on experience.

Nice to have

• Labeling tools such as CVAT, Label Studio or Roboflow.
• Object detection or segmentation formats (YOLO, COCO).
• OpenCV, NumPy, pandas, or SQL.
• Having trained a model yourself, even a small one.
• German.

Not a fit if

• You only want to train models and never look at the data.
• Repetitive, careful work for a few hours at a time is not for you.

How we work

• We spend time on site with our customers. You may visit a plant to see where the images come from.
• Facts over feelings: numbers with units, errors with next steps.
• You own your work end to end, with support from the ML team.
• Hybrid work from our Berlin office. Dogs are welcome.

How we hire

Apply with your CV. We read every application ourselves. Then a screening, a case study with real data, a session presenting it to the team, and a conversation with the founders. About three weeks from application to offer.