Geospatial AI & Computer Vision Expert

We’re exploring a Python-based tool for automatically extracting individual building and tent footprints from very-high-resolution satellite imagery. We’re looking for an experienced consultant to review the requirements and recommend a practical technical approach. Technical context - Satellite imagery: Pléiades Neo at 30 cm, SkySat and Pléiades at 50 cm. - Existing imagery and manually validated footprints potentially available for training, subject to a data-quality assessment. - Outputs: individual object polygons and counts, exportable as GeoJSON or shapefiles and compatible with ArcGIS. - Intended deployment: AWS g5.4xlarge with an NVIDIA A10G GPU, currently Windows-based. Linux is an option if justified. - Target processing volume: approximately 100 km² within a day. No model or architecture has been selected. We’re interested in an independent assessment grounded in practical experience. Relevant experience: - Hands-on experience with building or shelter extraction from satellite or aerial imagery, training and fine-tuning segmentation models, Python-based deep learning, geospatial data processing and deployment. - Experience producing usable GIS polygons from model predictions is particularly relevant. Please include a brief description of relevant projects, your contribution, availability and your rate for an initial consultation.