We are a well-funded stealth startup building a B2C platform to radically reform the U.S. residential solar industry. Our mission is to combat pervasive consumer distrust and inefficiencies by empowering homeowners with a transparent, high-fidelity, and interactive 3D solar design tool. We are a small, dedicated team working closely with an established industry partner for rapid, real-world testing and iteration.
We are looking for a Senior 2D/3D Computer Vision Engineer to build the core technology that makes our platform possible. The central challenge is to automatically generate clean, geometrically accurate, and simplified 3D roof models from raw aerial Digital Surface Models (DSM) and RGB data. Your work will be the foundation of our user experience, replacing the opaque and slow manual processes that dominate the solar industry today.
As the owner of this domain, you will ideally:
• Define the Process: Take charge of the entire ML pipeline, from R&D to production. This includes refining the annotation process and enhancing our existing Three.js-based 3D annotation tool to build a high-quality, large-scale dataset.
• Build the Models: Research, develop, and train a suite of models to handle roofprint segmentation, structured prediction of planar roof wireframes from point clouds, and detection/classification of roof obstructions (chimneys, pipes, windows, etc.).
• Ship to Production: Engineer a scalable and robust inference pipeline that can process new properties on demand.
Our stack is Python, PyTorch, FastAPI, Rasterio/GDAL, and JavaScript/Three.js for the annotation tool. We are looking for a pragmatic and experienced engineer who loves tackling complex 3D vision problems. You should be comfortable with both 2D and 3D data (especially point clouds and geospatial imagery) and have a proven track record of delivering end-to-end ML solutions.
To apply, please email vlad@design.solar.