fAIr

What is fAIr?

fAIr is an open-source AI-assisted mapping platform developed by the Humanitarian OpenStreetMap Team (HOT). The name carries a deliberate meaning:

• f — freedom and free and open-source software (FOSS)
• AI — Artificial Intelligence
• r — resilience and responsibility to communities

Using computer vision and machine learning, fAIr enables humanitarian mappers to detect features such as buildings, roads, waterways, and trees from open satellite and UAV imagery — contributing them back to OpenStreetMap for free use by anyone in the world.

The Problem

Low- and middle-income countries (LMICs) — the communities most in need of accurate maps for disaster response, development planning, and humanitarian coordination — are systematically excluded from the benefits of AI-powered geospatial tools. The reasons are structural:

• Commercial GeoAI platforms like ESRI ArcGIS and Google Earth Engine are proprietary, expensive, and inaccessible to local mappers and NGOs in LMICs
• Existing AI models are trained on data from high-income geographies, creating models that perform poorly or fail entirely in the contexts that matter most
• There is no open, community-governed space where humanitarian organizations and local mapping communities can collaborate on shared AI models
• Feedback loops that would allow communities to correct biases and improve model performance simply do not exist in closed systems

The result: the places most urgently in need of quality maps are the least able to produce them at scale.

fAIr 3.0 Vision — A Platform for GeoAI Models for Social Good

fAIr 3.0 represents a strategic evolution from a mapping tool to a platform for GeoAI models for social good — the open humanitarian equivalent of what Hugging Face is for general AI. The vision: a shared infrastructure where the global open mapping community and AI/ML developers collaborate to build, fine-tune, share, and deploy geospatial AI models grounded in local knowledge and community needs.

This platform connects five interdependent pillars:

My Role & Contributions

• Product vision and strategy: Defined and led the evolution of fAIr from a standalone mapping tool to the fAIr 3.0 platform vision — a shared GeoAI infrastructure for the humanitarian and open mapping community
• Roadmap and prioritization: Owned the product roadmap, translated community needs and field realities into technical priorities, and facilitated decision-making across engineering, design, and community teams
• Ethical AI leadership: Ensured the tool's development remained grounded in the needs of the communities HOT serves — applying principles of responsible, community-controlled AI throughout, including human-in-the-loop validation, bias mitigation, and local model governance
• Stakeholder and partner communication: Communicated AI/ML product strategy and progress to internal teams, external partners, funders, and the broader OSM community
• Open-source ecosystem building: Identified and supported complementary AI initiatives in the open mapping space, and contributed to positioning fAIr as a community-owned alternative to proprietary GeoAI platforms
• International representation: Represented HOT and fAIr at the AI Action Summit organized by the Paris Peace Forum, advocating for open, ethical AI for humanitarian good on the global stage

Open Source by Design

fAIr is licensed under AGPL-3.0 — one of the strongest copyleft open-source licenses — ensuring that any organization building on fAIr must contribute back to the commons. The platform is fully available for any community, NGO, or government to deploy, adapt, and build upon.

On GitHub, the project has grown to 132 stars, 79 forks, and over 2,400 commits, built primarily in TypeScript (frontend) and Python (backend and ML). The community spans individual contributors, HOT staff, and partner organizations across the globe.

Technologies

Links