In the first hours after a disaster, finding survivors trapped under rubble or cut off by floodwaters is the single most important factor in saving lives. Although new sensors and mapping tools exist, rescue teams still rely on proven methods such as trained dogs because the current technology often fails when infrastructure is damaged and time is scarce. RescueMap‑AI is a lightweight, scalable, open‑source system that combines geospatial data, crowdsourced reports, and machine‑learning‑based clustering to give responders real‑time, actionable maps.

Many field teams still use manual GPS logging, static satellite images, or drone fly‑overs. These approaches need stable power, high‑bandwidth links, and specialist staff, resources that are usually absent in the chaotic environment that follows an earthquake or hurricane. Conventional AI platforms also assume cloud access and powerful GPUs. RescueMap‑AI is designed for laptops, tablets, and rugged field computers, and it works even when the network is offline.

The platform has three main components. First, a simple input layer allows citizens, volunteers, and responders to submit reports through a mobile app, a browser form, or SMS in low‑bandwidth areas. Second, an integration layer merges these reports with freely available geospatial data, such as OpenStreetMap extracts, cached satellite tiles, or publicly released aerial imagery. Third, a processing layer applies a fast density‑based clustering algorithm that highlights areas where several distress signals overlap, creating an instant priority list for rescue deployment.

RescueMap‑AI follows an offline‑first design. Teams start by caching the base map for their region. After that, the entire workflow, including clustering and visualization, runs locally. When connectivity returns, the system synchronizes new reports with a central server so that headquarters can see the evolving situation. This approach bridges the gap between disconnected field teams and command centers.

During the talk, I will walk the audience through a simulated earthquake response in an urban neighborhood. Volunteers report sightings of trapped people through the app. RescueMap‑AI ingests these points, groups them by proximity, and displays “hot spots” on a web map that refreshes every few seconds. Field commanders can export the hot spots as GeoJSON or CSV to load into navigation units or print on paper.

The presentation will also cover present limitations and the roadmap. Current work focuses on improving location accuracy when GPS is degraded, authenticating user reports to reduce noise, and integrating drone imagery for rapid updates of blocked roads.

RescueMap‑AI is planned to be fully open source. The code base is written in Python and JavaScript with dependencies kept minimal. Contributions from the FOSS4G community are welcome, especially testing in varied terrains and translations of the user interface. By lowering the technical barrier, the project aims to extend advanced rescue mapping to volunteer groups, rural fire departments, and NGOs that cannot afford proprietary tools.

Attendees will learn practical methods for fusing open geospatial data with compact machine‑learning models, tips for building resilient offline applications, and ideas for mobilizing local communities through crowdsourced mapping. The talk will also outline possible pathways for integrating open‑source rescue technology into national preparedness programs, an aspect that supports my broader research on how software can reduce disaster mortality.

RescueMap‑AI shows that effective geo‑AI does not require expensive hardware or closed ecosystems. It only needs the right balance of open data, simple algorithms, and thoughtful design. I invite conference participants to explore the software, propose new modules, and collaborate on future deployments so that the next time disaster strikes, responders will reach victims faster.