At night, the road becomes a different product. Paint fades. Contrast collapses. A pedestrian can be present, moving, and obvious to a thermal sensor while remaining only a weak arrangement of pixels to a visible-light camera. The software in an autonomous system may be brilliant, but it cannot reason about information its sensors never collected. Chuck Gershman has built his latest company around that stubborn fact.
Gershman is the president, CEO, and co-founder of Owl Autonomous Imaging, a Fairport, New York company developing high-resolution long-wave infrared sensors and the perception software that interprets them. The system is designed to do more than produce a ghostly heat picture. It works toward classification, location, range, tracking, and scene understanding: the practical grammar a machine needs before it can act.
His pitch is less about replacing every other sensor than refusing to make one modality carry the whole job. Visible cameras provide color and detail. Radar supplies resilient detection and range. LiDAR adds geometric depth. Thermal imaging contributes information that does not depend on ambient light. Sensor fusion turns their disagreements and overlaps into a more complete account of the world.
“Is there something out there? What is it? Where is it? Where’s it going?”Chuck Gershman on the four jobs of perception
Those four questions sound almost elementary. They are not. A sensor that merely detects a warm shape has not classified it. A system that classifies a pedestrian has not necessarily measured the distance. A range reading does not reveal direction or speed. The machine must build an answer under time pressure, in motion, with imperfect inputs. In Gershman's telling, the point of a diverse sensor suite is that no single component should become a brittle point of belief.
A career measured in architectures
Gershman's route to thermal perception begins with packets, not pictures. In 2000, he was vice president of marketing and sales at Bay Microsystems as the startup prepared a network processor designed to handle a mix of carrier services. Two years later, as senior vice president, he was talking publicly about the company's first samples of a 10-gigabit processor. The vocabulary was different, but the operating challenge was familiar: take a difficult chip architecture and make its advantage legible to a buyer.
By 2006, he was Bay's president and CEO. An EE Times awards profile credited the company with 511 percent revenue growth and profitability in the prior year. Gershman named diversification as the next problem. The detail is revealing. Growth had arrived, but concentration could still make it fragile. The job was not to admire the curve. It was to ask what could break it.
He led Bay through two technology recessions and from its pre-revenue phase to more than $100 million in aggregate sales. Industry groups later recognized the company for growth and as a respected private semiconductor business. His public biographies also credit him with helping lead company exits involving Intel and PMC-Sierra. It is the resume of someone who learned that a technically correct product is only the opening argument.
Later roles broadened the canvas. He took the CEO job at Swedish networking startup PacketArc as it shifted its headquarters to the United States. He co-founded Kuveda, an analytics company, and worked on connected-device security at chip-design firm Intrinsix. Across public bios, his functions span engineering, marketing, sales, business development, operations, boards, fundraising, and strategic advising.
A former colleague captured the connective tissue: Gershman could boil advanced technology down to the elements that mattered to companies building products. That is not simplification as decoration. In deep tech, translation decides whether an architecture remains a diagram or becomes a business.
The thermal turn
Owl's technical roots came from work created under a US Air Force challenge grant involving the tracking of ballistic missiles in flight. Gershman and co-founder Gene Petilli saw a broader opportunity in the related thermal-ranging technology. Owl was formed in 2018, with a founding thesis centered on automotive safety and the places where ordinary cameras become unreliable.
“I am a semiconductor guy at heart,” Gershman said in an early company interview. The phrase matters because Owl's bet is not simply on a thermal camera. It is on a digitally architected focal plane that pushes more work into the digital domain, coupled with software for interpretation. The aim is to improve power, calibration, manufacturing, frame rate, and scale, then turn the sensor output into something an autonomy stack can use.
The comparison is not a winner-take-all scoreboard. Each sensor has distinct strengths and constraints. Thermal imaging can preserve meaningful contrast without visible illumination. Radar is robust and measures range but offers less spatial detail. LiDAR produces useful depth but brings cost, power, weather, and scaling considerations. Visible cameras remain information-rich when conditions cooperate. A safety system can use the diversity itself.
Owl calls its hardware family KnightOwl and its perception software KnightVision. The company has described megapixel-class digital thermal focal-plane designs and AI functions for ranging, classification, localization, and tracking. Its target is dense perception at a size, weight, power, and cost that fits platforms far smaller and more numerous than traditional premium thermal systems.
A technical core can stay fixed while the route to market changes.
The pivot without the costume change
Automotive qualification takes time. A supplier can have a credible technology and still wait years between a product commitment and production revenue. Owl began looking for mobility applications that could occupy that gap. Drones, robotic systems, industrial autonomy, perimeter sensing, and defense all encounter the same underlying problem: a machine must perceive reliably outside a controlled demo.
This is where Gershman's operating philosophy meets his technical one. On a 2026 podcast, he emphasized “time to decision” and the capacity to pivot. Investors, he argued, often care more about a team's adaptability than the permanence of its first plan. The company did not need to dress up as an entirely different business. It needed to notice where the same thermal architecture solved an urgent problem with a shorter path to use.
The new markets add harsher constraints. A small aerial platform cannot carry a sensor designed for a large vehicle. Power consumption becomes flight time. Weight becomes payload. Unit cost matters when systems are deployed in quantity. Manufacturing readiness is therefore not a late operational detail. It is part of the product.
In a June 2026 interview, Gershman described Owl's immediate work as advancing its next-generation thermal roadmap, scaling manufacturing readiness, expanding customer engagements, and pushing deployable AI perception forward. He also said the company was preparing to execute a significant US government program, without disclosing specifics. The language had moved from proving a modality toward making it repeatable.
“Leadership is about inspiration. It’s about vision.”Gershman on the difference between leadership and management
The steady hand behind the sensor
Gershman distinguishes management from leadership. Management can improve operations. Leadership creates a shared destination that employees, executives, and investors choose to follow. His version is not mystical. It sits beside decision speed, emotional steadiness, mentorship, and the willingness to update a plan without making every change feel like a crisis.
That stance has a practical history. Semiconductor companies live on long feedback loops. Capital arrives before certainty. Hardware choices become expensive commitments. External cycles can rearrange the market before the chip is ready. A leader has to maintain enough conviction for the team to keep building and enough flexibility to notice when the original commercial map is aging badly.
There is also a collaborative streak in the way he tells Owl's origin. In 2019, he praised the Silicon Catalyst accelerator not only for endorsement and industry credibility, but for surrounding the founders with experienced people. “It is fool hearty to think you can go it alone,” he said. The spelling may be informal; the operating idea is exact. A difficult company needs an unusually dense network of judgment.
Public recognition followed the earlier chapters: induction into Drexel University's College of Engineering Alumni Circle of Distinction in 2010, an EE Times executive award finalist spot, and four US patents. Three cover network-processor architecture. The fourth, issued in 2022, concerns multi-aperture ranging devices and methods. The list reads like a compressed record of his technical migration from moving data through networks to extracting distance from images.
What the machine learns at midnight
Physical AI is a fashionable label for an old collision: software meets weather. The world does not promise clean inputs, familiar objects, or helpful lighting. Gershman's work is a reminder that an autonomous system's intelligence begins at the edge, where photons, heat, silicon, lenses, power budgets, and manufacturing tolerances decide what information exists at all.
Owl still has to turn technical claims and customer engagements into sustained deployment. That is the ordinary burden of a deep-tech company, intensified by long qualification cycles and demanding buyers. Gershman is not new to the interval between an elegant architecture and a durable business. Much of his career has been spent inside it.
His aspiration is now clear enough to draw in one line: move thermal perception from a specialized capability into a scalable layer for autonomous machines. The opportunity begins after sunset, but it does not end there. Glare, smoke, dust, fog, and other degraded scenes all test whether a machine's view of reality is diverse enough to trust.
The useful image is not a car gliding through a perfect city. It is a small system operating at the edge of its power budget, at night, in weather nobody scheduled, still answering four questions in time. Something is out there. The machine knows what it is, where it is, and where it is going. Then the intelligence gets its turn.