RetroFin AI - team Sur
Team consisting of an iOS engineer, a PhD policy researcher from UPM, and an impact investment manager from iGravity with finance degrees from LBS.
Project Description
RetroFin AI helps Spanish homeowner communities navigate retrofit financing end to end (from uploading their building accounts to getting a signed loan application ready).We built this because the process is genuinely broken: communities have to deal with architects, contractors, banks, and subsidy applications separately, and most community presidents have no idea where to start and navigate burocracy and documents.
The agent pulls live Euribor via viaNexus MCP, stacks available subsidies (For example, plan Estatal 2026-2030 + Generalitat Catalunya in this demo), scores the community’s financial readiness, collects contractor quotes and flags spec issues, and generates the legal documents needed to apply. All in one flow a non-expert can actually follow.
We used Claude Sonnet for document extraction and financial summaries, FastAPI on Scaleway for the backend, FINOS archetypes.json to classify the building type, and React for the frontend. Financial calculations are deterministic Python . The 4 agent primitives we built (profile, validate, propose, apply) are common to structured finance workflows so the architecture is reusable beyond retrofit, particularly for complex financing processes.
Live demo: http://212.47.238.61:8000/static/demo.html
GitHub: https://github.com/SurRetrofinAi/retrofin-ai