XPRIZE Wildfire’s independent Judging Panel has chosen 11 teams to advance to the final round of the XPRIZE Wildfire Space-Based Detection & Intelligence track, following semifinal technical verification. The Judging Panel is awarding a total of $300,000 between six of the advancing teams in recognition of the distinct data analytic capabilities their systems demonstrated during Semifinal testing.
XPRIZE Wildfire Space-Based Detection & Intelligence Track semifinal testing was held remotely with teams testing from their home bases around the world. XPRIZE staff and Judges—experts in fire management, satellite intelligence, remote sensing, and other relevant fields—gathered at the Esri headquarters in Redlands, CA throughout the testing and judging periods.
Semifinals testing tasked teams with accurately and precisely detecting and characterizing target fires using historical fire data while minimizing false positives. Over the course of 12-hour testing windows, teams presented their diverse approaches, ranging from physics-informed modeling to deep learning and large language models, pioneered exclusively for detecting and monitoring wildfires in real-time. These cutting-edge solutions demonstrated the analytical strengths of each team's system and their ability to apply the necessary tools, methods, and procedures to process, arrange, and scrutinize Earth observation data effectively.
Congratulations to the teams as they advance in the competition! Their solutions underscore the immense potential of these technologies to help communities better prepare for natural disasters.
XPRIZE Wildfire is offered in partnership with Co-Title Sponsors Gordon and Betty Moore Foundation and Pacific Gas & Electric, Presenting Sponsor Minderoo Foundation, Bonus Prize Sponsor Lockheed Martin, and Supporting Sponsors Conrad N. Hilton Foundation, American Family Insurance, Fairfax Financial Holdings, the Roddenberry Foundation, Costa Navarino, and individual benefactors.
MEET THE TEAMS ADVANCING TO FINALS
MILESTONE AWARDEES
DEEPFIRE | BATON ROUGE, LA, UNITED STATES
DeepFire’s system is a combination of an AI-based wildfire prediction system and an AI-based wildfire detection system that interact synergistically. The detection system uses predictive risk maps to increase or decrease confidence in detections it is unsure of. The detector can forward information on current fires to a predictive model to predict future behavior, such as danger level, speed, direction of spread, etc. Supratik Mukhopadhyay, Professor of AI at LSU, leads the team. Previously, he led a team to the semifinal of the IBM Watson AI XPRIZE. Other team members include Dr. Robert Dibiano, Richard Barbalace, Dylan Wichman, and Matt Braun.
EMBER GUARD - A WILDFIRE WATCH GROUP | PALO ALTO, CA, UNITED STATES
Ember Guard’s approach fuses low- to high-resolution satellite data with atmospheric conditions and social media. Using deep learning along with LLM and AI techniques to detect wildfires, a scalable cloud-based high-resolution wildfire model forecasts the likely propagation and intensity of a given wildfire so that firefighters can prioritize their limited resources. Ember Guard is a cross-disciplinary team led by a high school sophomore along with materials scientists, entrepreneurs, wildfire-focused atmospheric modeling experts, mathematicians, and senior research scientists.
GENERATIVE INTELLIGENCE | MIJAS, SPAIN
Generative Intelligence leverages a cascading AI system to first identify high-probability fire zones by integrating satellite imagery and historical, climatic, and terrain data. High-risk areas undergo detailed analysis using cutting-edge AI to pinpoint fires, monitor spread, and evaluate impact, enabling swift emergency response. This scalable and adaptable solution can incorporate new data sources and expand to new areas. Generative Intelligence joins forces with top satellite companies to revolutionize firefighting. With cutting-edge AI and a personal stake in fire-prone Andalusia, Generative Intelligence is driven to save lives and land.
MAYDAY AI (GUARDIAN SPACE)
MAYDAY AI (GUARDIAN SPACE) brings together a unique blend of advanced artificial intelligence expertise and unparalleled satellite service capabilities to address the critical challenge of wildfires. Generative Intelligence leverages a cascading AI system, first identifying high-probability fire zones by integrating satellite imagery and historical, climatic, and terrain data.
MYRADAR | ORLANDO, FL, UNITED STATES
MyRadar specializes in building AI technology and robust data pipelines to deliver data, visualizations, and environmental alerts. Their patented miniaturized satellite technology uses spectral imagers and onboard AI optimizations to enable rapid alerting for environmental hazards, such as wildfires. The team’s mission is to democratize real-time information that allows better, more risk-informed decisions, and a constellation of satellites will provide data and alerting toward this goal. MyRadar is a technology company that builds situational awareness products trusted by millions of customers. The satellite team consists of several dedicated engineers and subject matter experts.
SIRIUS WILDFIRE ALLIANCE | LONDON, UNITED KINGDOM
SIRIUS Wildfire Alliance will deliver high-resolution wildfire nowcasting and forecasting by integrating Earth Observations, AI, GIS, a cutting-edge wildfire spread model and a unique combination of ground and space-edge computing. By leveraging onboard computations and efficient telemetry, SIRIUS can employ advanced but typically time-consuming techniques for granular spread and false positive detection, provide early information for emergency response, and recommend efficient firefighting resource allocations. SIRIUS brings together young engineers, established academics, researchers in top universities, ambitious startups from the UK and Australia, and wildfire management professionals.
ADVANCING TEAMS
FIRE EYE | HILLSBORO, OR, UNITED STATES
Fire Eye employs multiply-sourced hyperspectral image analysis within existing infrastructure for cloud computation and ground control stations. The system uses ensemble models to predict fires and their behavior. Multiple communication channels for effective information dispersal. Team Fire Eye is a veteran in first responder-based research and development with extensive expertise in computational vision algorithm development, aerospace sciences, and communication protocols.
ORBITAL SIDEKICK | SAN FRANCISCO, CA, UNITED STATES
Orbital Sidekick will use a combination of hyperspectral and multispectral satellite images which allow speciation of vegetative materials and detection of combustion. Orbital Sidekick’s GHOSt constellation currently consists of six satellites, five of which are already in orbit and acquiring data, with seven in development. Orbital Sidekick is building the most advanced constellation of hyperspectral satellites with unmatched global monitoring capacity through its spatial and spectral resolution. Orbital Sidekick’s XPRIZE team consists of hyperspectral imaging and data scientists as well as payload engineers, technical leaders, product managers, and UI/UX designers.
REDBACK FIRE TEAM | MELBOURNE, AUSTRALIA
Redback Fire Team proposes a two-phase, integrated and robust solution to wildfire surveillance and characterization. Phase 1 aims to detect wildfires within one minute using a constellation of EO sensors. The algorithm is tailored to accommodate geographical, seasonal, and diurnal variations. This module is currently operational in Australia, delivering hotspot notifications within 20-45 seconds. Phase 2 accurately characterizes fire detections and behavior using a hybrid-model approach within 10 minutes. Redback Fire Team is composed of members from RMIT University & Covey Associates PTY LTD, bringing together expertise in wildfire detection and attribution, and fire behavior & modeling.
SNUFFED | FLAGSTAFF, AZ, UNITED STATES
Snuffed proposes a “string of pearls” constellation of 90 micro-satellites or CubeSats in low Earth orbit (LEO). The orbital period in LEO is 90 minutes, so one unit in Snuffed’s constellation passes over a given ground location once per minute. The team’s rapid detection strategy, when fully implemented, has the potential to reduce these costs and damages enormously. This is a game-changing advance in humanity’s approach to combating wildfires. The Snuffed XPRIZE Wildfire Team currently has 25 engineers, scientists, firefighting and fire management experts, students, and private sector industry professionals.
WOOLPERT DI CLOUD GEO TEAM | DAYTON, OH, UNITED STATES
The Woolpert DI solution is based on the orchestration of several models and satellite sources through different phases: early risk assessment, rapid fire detection, fire characterization, and operational delivery. Forecasting models, computer vision, deep learning, and first-principles models will be deployed alongside a variety of public and commercial satellite sources. The overall workflow will be automated into a platform designed to send operational insights and alerts to relevant stakeholders. Woolpert DI is a team of remote sensing scientists, cloud engineers, AI/ML experts, and environmental scientists that build scalable models and analytical platforms based on Earth Observations and AI.