Satellites on Fire: teen-built AI spots wildfires 35 min before NASA; $2.7M seed

Wildfire early-detection service: a software platform combining 10+ satellites, tower cameras and weather data with three AI models, alerting fire services and companies in real time by WhatsApp, SMS or email

That wildfire response needs no new hardware — a software layer fusing others' satellites, cameras and AI can detect fires faster than NASA

US$2.7M seed (April 2026) led by Dalus Capital with Draper Associates, Vitamin C and Draper Cygnus; ~US$900k raised earlier from backers including the UN, MIT, Cornell and Alexis Ohanian

In 2020, wildfires around Córdoba burned 340,000 hectares and destroyed homes of people close to Franco Rodriguez Viau, then a 16-year-old at ORT technical school in Buenos Aires. With classmates Joaquín Chamo and Ulises López Pacholczak he built a prototype that tapped NASA satellite feeds; fire agencies told them alerts arriving three to five days late were useless, so the team restarted around early detection.

Rather than build its own satellites or cameras, the team made a software platform that fuses data from more than ten satellites (NASA, NOAA, ESA), 360° tower cameras, drones and weather forecasts, running three AI models: ignition detection, smoke-column recognition and spread prediction. Alerts go out by WhatsApp, SMS or email, and user reports feed more than 20,000 field validations that retrain the models. The first paying customer came about 18 months in; clients grew to forestry and energy companies, Argentina's volunteer firefighters and insurer AON, and in 2025 its alerts supported responses to more than 600 wildfires.

The wedge was timing, not sensing: fires were already visible to satellites, but alerts landed three to five days late, so crews could not act on them

Software over rented hardware scaled where satellite builders could not: no capital-heavy assets, faster into 21 countries

The free tier was a data flywheel: 20,000+ user field validations made its models detect fires faster and more precisely

Buyers had a real budget line — fire loss — and the platform sold to insurers, forestry, energy and agriculture, not just governments

You can beat a vertically integrated incumbent without owning the asset: a software layer over existing satellites beat NASA's alert latency, and the free tier fed the field data that sharpened the AI

As of May 2026 the startup had 55,000+ users in 21 countries, 33 active clients and a 26-person team, and had just won the UNCCD G20 Youth Ecopreneur Award at ChangeNOW (31 March 2026). It plans to spend the seed on better AI models, CO2-avoidance measurement tools and a parametric wildfire insurance product with AON, and to expand from Latin America into the United States; the long-term goal is drones that verify and eventually fight fires