brings 'computer vision intelligence' to physical security

Jan. 19, 2022
Company exits stealth mode with $52M already raised in venture funding

It seems the security industry is awash in a plethora of new video analysis solutions that are rapidly making up for the shortcomings of prior technologies. Indeed, today’s artificial intelligence (AI) platforms, which are powered by ever-evolving machine learning and deep learning software solutions, are finally providing the capabilities and return-on-investment that systems integrators and end-users were clamoring for over a decade ago.

The technology has become so ubiquitous that it is becoming increasingly difficult for AI developers to stand out in the crowd. However, a new entrant to the market,, which emerged from stealth mode on Wednesday, is looking to differentiate itself by being able to provide intelligent video analysis at scale.

The company was founded in 2017 by two Silicon Valley veterans – CEO Shikhar Shrestha and CTO Vikesh Khanna – who have spent the last four years developing their solution dubbed “computer vision intelligence” or CVI, for short. CVI can leverage existing camera hardware to analyze and contextualize video data to automatically detect security incidents while simultaneously reducing false alarms and preserving privacy.

According to Shrestha, one of the things they noticed when they started peeling back the onion of the security operations in most current enterprise and campus environments is that they remain primarily reactive in nature and require an inordinate amount of manpower to sift through and determine which alarm events require a response.   

“If you look at how [the industry] does security operations or how we secure a building or campus environment currently, at a high level, what we are doing is basically three things: the first thing we’ll do is go in and install hundreds of security cameras, thousands in some cases, all around the site that nobody is actually watching,” Shrestha explains. “The second thing we’ll do is setup different kinds of alarms – PACS, door contact sensors, glass break sensors, and all of these different end points that generate alarm events, and they end up generating hundreds of thousands of alerts every, single day and 99% of them are nuisance alarms. And the third thing we do is we have security officers, which are great from a response perspective, but the challenge is you have a couple of officers at a large site patrolling, and we have them do observe and report; the problem is they can’t be everywhere at the same time.”

Contextual Analysis at Scale

Shrestha says Ambient, which has already raised $52 million in venture funding, essentially acts as an “automated brain” in the middle of all this, taking in all of the sensor and camera data and ferreting out real incidents from false alarms to dispatch security personnel as necessary.   

“It retrofits into these existing cameras and alarm sources and basically uses computer vision intelligence to mine all of the camera feeds and alarm events for a whole range of threats and suspicious events. We call them ‘threat signatures’ basically and we look for a hundred of them in the product today and when we detect one of those events, we send a real time alert so the security team can respond and actually prevent the incident from happening,” Shrestha adds.

Shrestha, who became the victim of an armed robbery when he was 12 years old, says that helping to develop this solution and advance security technology more broadly was a deeply personal endeavor.

“I grew up tinkering and building security alarm systems, burglar alarms and intrusion detection systems and was super paranoid. I am always concerned about security,” he explains. “I also started realizing over time that the victims that are involved in security incidents, the way things happen today, they’re always looking at cameras and expecting that someone is watching, but that’s just not the case. The way our industry works is completely reactive.”  

What differentiates Ambient from other AI video analysis solutions on the market, according to Shrestha, is the ability to do contextual analysis at scale as opposed to simple object classification.

“All of the signatures we look for are sort of like contextual patterns of suspicious behavior or indicators of compromise, like a person that tailgated into the lobby, started running and it is afterhours, so the reception desk is unoccupied,” he says. “It is a materially different technology from a computer vision perspective from what we did before with analytics versus what is possible to now be able to really understand the context of the events that happened.”

Building the Channel

Even as a stealth company, Shrestha says that Ambient has already achieved a great deal of success with a number of large tech and Fortune 500 organizations, such as Adobe and VMware, having adopted their technology and realized operational ROI. However, moving forward, the company is going to be focused on growing their channel partnerships and bringing the technology to a diverse range of customer applications.

“Going forward while we’re coming out of stealth, we want to be able to serve customers across more verticals and more geographies and bring the platform to basically all of these security operations and deliver those benefits,” Shrestha says. “Building the channel is a big priority for us. We already work with a lot of the regional VARs and systems integrators as well as the nationals and our approach right now has been to collaborate with them – make the customers successful, help the channel partner learn how to communicate the value proposition with the product, what does the implementation look like, and then sort of cement that into a program. Our product is subscription software, so we can generate RMR for the integrators and there are opportunities in the implementation itself for setting up the system.”   

Joel Griffin is the Editor-in-Chief of and a veteran security journalist. You can reach him at [email protected]