This Manayunk company creates AI surveillance for 100 U.S. buildings, part of an evolving Philly start-up scene

The people who built 3.0 University Place, an eight-story lab and office complex that covers a landscaped block of West Philly, had a problem: During the biotech slump after their 2023 opening, security costs escalated faster than occupancy.

“We had 80 Siemens cameras in there,” remotely monitored from a suburban surveillance office, and were “paying more than $150,000 a year for guard shifts,“ said Anthony Maher, the University Place Associates president.

But the building’s New York owners were still getting midnight calls for nonevents. Meanwhile the human video watchers missed incidents of concern — the stranger who had taken to sleeping on a tenant’s canopy at night, the fired contractor who sneaked back into the building one weekend.

Investors pressed them for a better way.

Maher and his partner, Scott Mazo, went hunting for an artificial-intelligence solution — an AI-era automated security system to collect and analyze camera, physical, and human-observed surveillance data; rank threats; sort out false alarms; open or shut locks; and quickly notify law enforcement, emergency services, management and owners’ groups as needed.

They rang an entrepreneur they’d met five years earlier at a Drexel University pitch meeting: Matias Klein, founder of Manayunk-based Kognition AI.

Kognition specializes in what Klein calls “cyber-physical threat detection systems,” which link and analyze digital security video, activate remote locks and sensors, load key data into formulas matching clients’ priorities and requests, sort and share communications to alert the right people and agencies, and update responses based on results.

“You hang the cameras; we network the building systems,” as Klein puts it.

At 3.0 University Place, “Matias’ system gave us full access to the best 25 cameras, every floor, all times,” said Maher. The system caught incidents, routed them to the right people, and suggested simpler warning paths.

“They keep teaching us to manage the building, telling us, for example, about our people coming in when they didn’t really need to,” he said.

It worked so well “we told the owners to take the security-guard line-item out of the budget,” though they still use guards on occasion. It helps that the building’s big new tenant is the city police forensics unit: “We have a lot of cops coming here,” Mazo added, laughing.

More than 100 North American properties use Kognition AI, from Scotia Plaza, one of the largest office buildings in Toronto; to Pa.-based Benco Dental locations in several states.

Kognition, founded by Klein in 2017, was one of a handful of Philadelphia-area AI start-ups that presented to local investors at a June gathering. Investors included Osage Capital, MissionOG, and Susquehanna International Group’s SIG Growth Equity Fund.

“The AI landscape is changing so fast, it’s hard to tell what will be supplanted in a few months,” said Ellen Weber, who runs Robin Hood Ventures, and attended the meeting.

Robin Hood is a Philadelphia investors’ group whose members include start-up veterans. Members’ portfolio of local AI-dependent start-ups include Center City-based Deepwave, which makes laser weapons guidance systems for Air National Guard jets and other military projects; University City-based GreenIRR, which uses AI to speed carbon-emissions detection for truckers; and Proscia, which digitizes biotech images for fast diagnosis and pharmaceutical drug trials.

Thomas P. Dwyer, a partner at the law firm Troutman Pepper Locke who organized the June meeting, said Philadelphia trails other AI hotbeds.

“Philly has a lot of potential and opportunity here, especially if some of its larger companies get involved, as they do in Silicon Valley,” he said.

AI, he noted, is still seen as a threat by many local companies. They worry, “Can someone armed with Claude overtake your business, duplicate your software, and compete with you immediately — without having to hire people?”

And veteran investors are “gun shy” about AI profitability, he said.

Which, he added, leaves opportunity to those investors who make the “leap of faith” to promising AI start-ups

“An AI product makes itself better” the more it is used, said Kognition AI founder Klein.

Some of the best-known (and most-feared) AI image-and-data applications are based on giant AI companies’ mass sorting of millions of individuals’ information into ever-improving, readily searchable central databases at big new data centers — raising familiar privacy worries, as seen in the recent controversy over the alleged misuse of Flock digital camera data that police have used in attempting to quickly identify drivers.

By contrast, Kognition is one of the firms that operates at the “edge” of big networks, Klein said in an interview at his firm’s Manayunk meeting space. “We don’t send customers’ data throughout our network, and we don’t preload that information,” he said. “We don’t flow data back to a central monitoring station.”

Rather, the goal is “to give people control over the spaces where they live or work or worship, so they can get into a defensive posture fast, if they ever need to.” Custom, localized, focused solutions are “a big area of value creation” for AI start-ups, according to Klein.

Kognition AI’s custom sharing limits and options are part of its appeal to Ravi Kalidindi. He was tasked by leaders of Bharatiya Temple, a Jain and Hindu religious congregation in Montgomery County, with setting up a system “that gave us control of our doors” without intimidating visitors or compromising members’ private information.

Kalidindi turned to Kognition as a locally based provider.

The system at Bharatiya starts with security cameras at the main entrance and stairways, integrated into a software and communications platform. “They check if our doors are closed. They tell us where there is human traffic. We can let the cleaning crew in without a key. It has mobile device control. It can even work on fobs,” Kalidindi said.

Kognition also has capabilities the temple has not used: “It has a weapons-detection system we have not enabled. It has additional sensors that can ID what people are carrying. It can tell us if we are seeing an above-average or above-expected crowd, make a note of the faces of people who are entering for the first time, and trigger an alert.”

There is a sense of user control of the AI systems that Kalidindi finds reassuring: ”We want to enhance the sense of security, that our facilities are being monitored, but also that everyone with good intent is allowed to enjoy the premises."