Sonia Kastner, the Founder and CEO of Pano AI, has been creating waves in the field of wildfire early detection and intelligence.
With a background that spans over a decade in building and leading supply chain and manufacturing organizations in fast-growing, mission-driven startups, her expertise is extensive.
Pano AI, under Sonia Kastner’s leadership, stands as a leader in the domain of wildfire detection, bringing together a diverse team of experts in technology and business.
The company is focused on innovating technology solutions to aid fire professionals in combating the threat of wildfires and assisting communities in strengthening environmental sustainability and resilience.
In this exclusive interview with Kastner, FSJA Editor Iain Hoey delves into the groundbreaking world of AI-driven wildfire detection, exploring the innovative technologies and visionary strategies behind Pano AI’s success.
Throughout my career, I’ve worked in government, technology, and business, and have learned the transformative potential of each sector.
In 2007, I was led to the green energy movement during my MBA at Stanford, followed by a role at with Alta Devices, a solar energy start-up, before the sector’s decline in Silicon Valley.
Later, my dive into the Internet of Things included the fast-growing, mission-driven company Nest, where I worked on one of the first, security cameras utilizing AI.
These experiences helped to pave the way for what would become our first product, Pano Rapid Detect.
By 2019, after witnessing devastating wildfires in San Francisco, including the Tubbs and Camp Fires, I wanted to do something to reduce the risk of these kinds of megafires from happening again.
Partnering with my co-founder, Arvind Satyam, we sought to leverage IoT to revolutionize fire response.
Our research revealed that the firefighting industry was hungry for technological advancement As wildfires intensified, the call for enhanced tools became evident.
Our vision aimed to equip firefighters with state-of-the-art technology, marrying our diverse experiences to address a global challenge, to improve fire fighter safety, and to better safeguard communities.
Pano Rapid Detect represents a leap in technology for fire professionals.
By leveraging advanced visual equipment and the power of artificial intelligence, it offers a platform that allows for the rapid detection, verification, and dissemination of crucial information regarding new ignitions.
Central to this system is the integration of myriad data streams: from ultra-high-definition cameras and satellite data to emergency alerts and weather data.
This integrative approach grants emergency managers a 360-degree perspective on potential fire threats, facilitating faster and more efficient containment efforts, ultimately helping to safeguard lives, habitats, and properties.
Our solution, Pano Rapid Detect, epitomises proactive fire detection.
Strategically placed ultra high-definition cameras on structures like cell towers and water tanks continually monitor the environment, hunting for the faintest hints of smoke.
This 360-surveillance camera is improved by additional data from 911 calls, satellites, and other sources.
When a potential wildfire is detected, the human analysts in the Pano Intelligence Center immediately tags and triggers alerts.
We use a proprietary methodology to very accurately triangulate fire location using multiple camera views.
First responders not only receive alerts rapidly but are also equipped with real-time footage of the fire’s progression, enhancing coordination and response time.
Pano is the only company to offer an end-to-end turnkey solution that offers customers fire intelligence as a service, leaving Pano to take care of the rest.
All under one roof, Pano designs the hardware, software, and AI, we manufacture the equipment in our own factory, and we are responsible for designing, deploying and maintaining the remote camera network as well.
Traditional fire detection methods, often reliant on 911 callers, are susceptible to false positives and delays.
In contrast, a Pano alert provides direct, real-time detection, supported by verified visual evidence.
When every second counts, Pano Rapid Detect helps to ensure that first responders are armed with accurate, timely information, enabling them to tackle the blaze with unparalleled efficiency.
Historically, staffed lookouts stationed on mountaintops helped detect fires.
However, as wildfire risk expands into the Wildland Urban Interface, it is often not feasible to place a staffed lookout tower on the highest hilltops because those hilltops are already occupied with structures, such as cell towers.
But it is feasible to mount a camera to those existing structures and use AI to perform the role of lookout.
At Pano AI, we leverage advanced computer vision models to actively detect new wildfire ignitions in real-time.
We use deep learning object detection AI, which relies on collecting data on wildfire smoke and other similar objects that could be confusing, like dust, fog, and clouds in order to label it as ‘smoke’ or ‘not smoke’.
While AI plays a pivotal role, human judgement remains invaluable.
Each detection that our AI system flags undergoes rigorous examination by human analysts at our Pano Intelligence Center.
This collaborative interplay of machine and human intelligence ensures the alerts we issue are actionable.
Pano AI’s blend of AI and human expertise ensures alerts are not mere notifications but urgent calls to action.
As wildfires escalate, our dual-layered approach offers a reassuring defence, making us better equipped to help first responders tackle such threats.
Wildfires, with their evolving dynamics influenced by terrain, vegetation, and weather, present a unique challenge for AI detection compared to static subjects like parked cars.
Our system at Pano is designed to continuously learn from this complexity.
Drawing from deployments in seven US states and multiple regions in Australia, we’ve amassed diverse wildfire data spanning snowy terrains to grassland infernos.
This extensive database aids our AI in predicting and understanding fires within their distinct environmental contexts.
Recognizing the vast differences in wildfires, such as those in Oregon’s dense forests versus Australia’s arid regions, underscores the importance of our diverse data set.
Our robust data-labelling process meticulously categorizes new information, reintegrating it into our training sets, ensuring our AI remains adaptive to the multifaceted and evolving nature of wildfires.
Though no system is flawless, Pano AI’s expansive training data, combined with our commitment to continuous learning, sets us at the forefront of detection capabilities.
As we gather more data from varied terrains, our proficiency in discerning wildfires only strengthens, ultimately aiming to better prepare communities and first responders against the erratic nature of wildfires.
