Its heartbreaking. Losing your home, your neighborhood, your memories, your relationships. A series of 14 destructive wildfires affected the Los Angeles metropolitan area in California in January 2025. The ensuing devastation killed at least 30 people, forced more than 200,000 to evacuate, destroyed more than 18,000 homes and structures, and burned over 57,000 acres of land in total. The costs and time to rebuild are staggering – in the $100Bs and ~ 10 years. The deaths and damage to property from these fires made them likely among the most destructive fires in California’s history
How do we tame these fires these before they spiral out of control?
Crewasis.ai, a startup out of TechStars L.A. accelerator, uses publicly available social media content with a proprietary data science and AI platform to predict and understand customer trends and sentiments. (Disclosure: I am an advisor to Crewasis.ai). They had previously used their platform to analyze consumer sentiments on the rapidly progressing revolution of autonomous (driverless) cars. For this article, they analyzed ~300K social media posts during the January-February 2025 period to reveal thoughts and recommendations from the public for managing future wild fires. These are listed below:
- Early detection using AI and sensor technology is crucial for preventing loss of life and property
- Incorporating fire resilience in future designs
- Establish trusted communication channels for fire preparedness and early evacuation
The first insight is the most effective and subject of Part 1 of this article (the sensing and AI solution). Part 2 will report user experience of this capability (Florida Forest Service and San Bernardino County, California Fire Protection District).
Fire Neural Network (FNN) is a venture backed company that germinated out of research in 2021 at the University of Florida. It was co-founded by Dr. Istvan Kereszy and Caroline Comeau. Wildfires can be caused by natural (electrical transmission failures and lightning) as well as human (arson, vehicle accident, etc.) events. In many parts of the United States, lightning is a dominant cause of natural wildfires (over 50% of such fires according to ESRI). Lightning creates a continuing current with durations ranging from tens to hundreds of milliseconds and exhibit amplitudes from several tens to a few hundreds of amperes. The core intellectual property that FNN has commercialized is the High Risk Lightning Detector (Figure 1) which detects these current profiles over a 20 mile radius and within 40 seconds of the strike.
FNN deploys a network of such detectors in wildfire prone areas for commercial and government customers in the United States as well as internationally. Real-time data from these sensor networks is combined with GIS (Geographical Information Systems) to integrate location specific weather and vegetation data into a trained AI (Artificial Intelligence) engine to assess wildfire risk due to lightning strikes (Figure 2). Time and accuracy are of the essence here – early detection (within 40 seconds) and notifications help get fire-fighting resources to the right locations to eliminate wildfires before they spin out of control.
Foreflight is FNN’s tablet-based wildfire alert application issued to fire fighting agencies along with high risk locations. This enables quick verification and containment of the highest risk wildfire sites before they spread and spiral out of control (Figure 3):
The State of Florida has one of the highest rates of lightning in the United States. This, combined with the amount of foliage and weather conditions makes lightning-related wildfires essentially an all-year phenomenon, with peak activity in spring (February-June) when the weather is dry. Fighting wildfires through use of technology is a state government priority. The non-profit XPRIZE Foundation (established in 1994) which has hosted competitions to address global challenges announced the XPRIZE Wildfire in 2021, a 4-year, $11 million competition is meant to incentivize the innovation of autonomous and time sensitive wildfire management technologies.
FNN is one of the finalists in this competition through the GatorX team which includes the University of Florida, NVIDIA, SATLANTIS (satellite imaging), SwissDrones, Phenix Systems and Archer (Drones), N5 and SWIR Vision (Sensors) to identify, locate, assess and eliminate wildfire germinations autonomously, without compromising human safety. The HiPerGator, a NVIDIA equipped AI computing platform established by NVIDIA co-founder Chris Malachowsky (a University of Florida alumnus) plays a critical role in absorbing the terabytes of video, climate, vegetation, communication, location and sensor data, and providing actionable information to drone-based, fire verification and mitigation platforms.
SwissDrones, based in Zurich, was founded in 2013 by ex-pilots and aviation experts to develop and sell drones to oil, gas, and utility companies to inspect pipelines and infrastructure in remote areas, as well as to civil authorities for search and rescue missions. The drones are VTOL (Vertical Takeoff and Landing), twin rotor autonomous helicopters powered with aviation turbine engines, and have a take-off weight of ~200 lb (fuel + payload), speeds of ~30 mph and mission range of ~50 miles (Figure 4). This enables BVLOS (Beyond Visual Line of Sight) operation to cover broad geographic regions for day and nightime operation. The drones are equipped with precision navigation and GPS sensors, and incorporate multiple payloads (as required by the customer and their application) including infrared and mid-wave (3-5 um wavelength) thermal cameras, and LiDAR imaging for 3D awareness and mapping. As a partner on the GatorX team, the role of the SwissDrone platform is to fly to the location indicated by the FNN ForeFlight application and confirm whether the fire threat is real. If it is, the information triggers firefighting action by specialized aerial drones and ground-based fire fighting teams.
Phenix Solutions, founded in 2011 by ex-pilots and aviation professionals is based in Oregon and “delivers game changing aircraft worldwide in order to save lives and maximize demand on command”. The company specializes in design and manufacture of heavy lift (above 3,000 lbs payload) Remotely Piloted Aircraft (RPA) for various civil, defense, firefighting, search/rescue, disaster management and construction applications. It is a member of the GatorX program with FNN and others. Once prompted by inspection drones like SwissDrones of the verification and location of wildfires, the Phenix drone deploys and commences mitigation actions (Figure 5).
Equipped with Rolls Royce engines, its VTOL drone-helicopters can carry and dispense 3000 lb containers of liquid firefighting solutions at targeted locations to extinguish fires before they spiral out of control. Their vehicles are FAA certified and used in firefighting and other hazardous applications to prevent loss of human life, and respond quickly to threats before they escalate. In current operation, a human pilot plans and controls the flight path. The eventual goal is to use AI and sensor data to achieve full autonomy.
The causes of wildfire can be due to human (intentional or unintentional) causes. But as the 2025 Los Angeles fires show, the impact is devastating. Using advanced sensors, imaging, AI and autonomy to locate, verify and suppress fires at their origination is critical and can return huge benefits. The operational process flow is complex but achievable. Part 2 of this article will report user experience of this capability (Florida Forest Service and San Bernardino County, California Fire Protection District).
