UrbanClimateAI
An applied AI project at Fraunhofer IAO for extracting climate and urban information from documents, linking it to locations and assets, and retaining references to the source material.
- Partner
- Fraunhofer IAO
- Period
- Jul 2026 – PresentOngoing
- Focus
- Climate Intelligence
Problem
Climate and urban information is distributed across municipal climate plans, flood-risk documentation, environmental reports, and planning documents. Connecting this information to individual assets requires document processing and spatial linking.
The project investigates AI/NLP methods for extracting and structuring this information while preserving source references. Intended users include banks, real-estate companies, retailers, and other organizations that own, finance, manage, or develop physical assets.
Contributions
- AI system and software architecture
- Backend and frontend development for Version 1
- Development of the initial MVP interface
- Technical study of document processing, spatial linking, knowledge retrieval, and validation
Tech Stack
Proposed in the technical study
- Python
- FastAPI
- PostgreSQL/PostGIS
Proposed for Phase 2
- PyMuPDF
- Mistral OCR
- LanceDB/Qdrant
- BM25 and vector search
- LLMs
Timeline
- July 2026: project preparation, architecture, and technical concept.
- July–August 2026: AI architecture and Version 1 implementation.
- August–September 2026: backend/frontend development and initial MVP interface during Sprint 1.
- September 2026: Stuttgart MVP validation and Phase 2 pitch.
- Current: further development, additional regional/climate data integration, and exploration of potential pilots.