This 6-month internship is part of osapiens Terra, a specialized team in Munich that combines remote sensing data and AI in sustainability software. We build enterprise software for deforestation-free supply chains. Our SaaS solution ensures compliance with the EU Deforestation Regulation (EUDR), which affects thousands of European companies and their suppliers. We track the supply chain end to end, from farmer to retailer, and use satellite data and deep learning to detect deforestation.
Our models run in production and are engineered to generalize across geographies, stay robust, and remain explainable. As an intern, you'll build and ship them with us.
- Train, fine-tune, and evaluate deep learning models on satellite imagery using PyTorch
- Evaluate the performance of existing models and maps, find where they fail, and fix it
- Read and implement recent computer vision and remote sensing papers, and turn them into working code
- Take prototype methods to production-ready code that scales across large geospatial pipelines
- Own a project end to end: from data preparation and training through evaluation and deployment
- Write clean, efficient, tested, and well-documented code in a shared codebase
- Share ideas for new models and algorithms, and explain your work clearly to the team
- Late-stage Bachelor's or Master's student in Computer Science, Electrical Engineering, or a related field
- Strong foundation in machine learning concepts and algorithms
- Hands-on experience building and training deep learning models in PyTorch (projects, thesis, research, or prior internships)
- Solid Python engineering skills: you write maintainable code, use Git, and are comfortable on Unix-based systems
- Ability to read research papers and turn them into implementations
- Excellent English and communication skills
- You work independently, take ownership, and are just as comfortable in a team
- Curiosity and a drive to learn new things quickly
Nice to have:
- Computer vision research experience
- Experience with satellite imagery or geospatial data (multispectral imagery, GeoTIFFs, GeoJSON, rasterio/GDAL, etc.)
- Experience scaling training or inference (GPUs, distributed training, cloud environments)
- A purpose-driven mission tackling complex sustainability challenges while working alongside global industry pioneers at a fast-growing unicorn company
- Real-world, unsolved problems in geospatial AI and remote sensing
- Hands-on experience across the full ML lifecycle: data, training, evaluation, and deployment
- Your work ships. If your model performs well enough, it goes into production and reaches 2,500+ customers
- A modern Munich office with flexible hybrid work options, open team culture, and flat hierarchies.