A serial entrepreneur from Tel Aviv now headquartered in a low office on Santa Cruz Ave, Los Gatos, where 50,000 cameras quietly report to a platform he built to make them think.
Founder & CEO - Lumana
Sagi Ben-Moshe spends his mornings running a company that watches things. Not in the creepy way. In the way a good security guard watches - the one who notices the unlocked back gate at 4 a.m. because something about the pixels is wrong. The difference is that Lumana, the company he founded in 2021 and unveiled from stealth in April 2024, does it across 50,000 lenses at once.
His pitch to a customer is short. Keep your cameras. Keep your hardware. Plug into Lumana's hybrid-cloud platform and your existing surveillance system stops being a tape you rewind after the fact and starts being a colleague that pings you in real time. The platform processes more than a million data points a second - people, vehicles, faces, license plates, behavior - and tags the events you actually care about. McDonald's is a customer. So is Meta. So is the Minnesota Twins Baseball Club, New York University, and a backyard-shed company called Tuff Shed. The list is delightfully unbothered by category.
That breadth is the point. Ben-Moshe is not selling a niche product to a single vertical. He is selling, in his words, "the infrastructure for cameras to perceive, understand context and take action, not just watch." It is the kind of sentence that sounds like a slogan until you remember he spent the previous decade trying to do this from inside Intel.
Lumana raised a $24 million seed in April 2024, led by Norwest Venture Partners and S Capital. In July 2025 the round behind it was a $40 million Series A. Total capital to date - $64 million. Headcount sits around 82 across offices in California and Tel Aviv. The company describes itself as a hybrid-cloud, camera-agnostic, NDAA-compliant AI video security platform, which is a lot of hyphens for a single sentence and exactly the point - they are determined not to be a hardware lock-in story.
What's interesting about Ben-Moshe is not that he is doing this. It is that he had to leave a very large institution to do this.
In 2012 Intel bought a 3D imaging startup he had founded. He stayed for nine more years. By the end of it, he was running RealSense - Intel's bet on giving everyday machines depth perception - and serving as Chief Incubation Officer of the Emerging Growth and Incubation group. Trade press at the time identified him as Intel's most senior Israeli executive.
Then, in August 2021, Intel announced it was winding RealSense down. Ben-Moshe left the same window. A division he had built for a decade was retired. He did not, judging by what happened next, take a long break. By the end of that year, Lumana existed.
The lesson he carried out was specific and unsentimental. Computer vision works. The constraint is not the math. The constraint is the surrounding system - the cameras, the storage, the deployment realities, the operator who has to live with the alerts. Lumana is, in many ways, an answer to the question of what RealSense could have looked like if it had been built as software for installed hardware rather than as a hardware product chasing developer adoption.
He has spent his entire working life on the problem of teaching machines to see. The companies change. The question doesn't.
Founds DGScreen and Invision L.T.D. in Israel - early work in 3D imaging and visual sensing.
His 3D imaging company is acquired by Intel. Joins as part of the deal.
Senior roles at Intel - VP and GM of RealSense, then Chief Incubation Officer and Corporate VP of Emerging Growth and Incubation.
Leaves Intel as the company shuts down RealSense. Founds Lumana the same year.
Lumana emerges from stealth with $24M seed led by Norwest and S Capital.
Series A - $40M raised to scale platform and go-to-market.
50,000 cameras under management. Customer roster spans retail, sport, education, hospitality, manufacturing, healthcare and government.
Lumana's marketing language is loud with the usual suspects - AI-powered, real-time, hybrid-cloud, edge processing, behavior-based detection, license plate recognition, agentic AI. Underneath the buzzwords the architecture is sensible. Heavy inference happens partly at the edge, on appliances installed near the cameras, and partly in the cloud where models can be retrained and dashboards can be built. The platform is camera-agnostic by design. If your customer already paid for a thousand domes, you do not need them to throw the domes out.
The output is the part that matters. Operators get alerts that mean something - a panic-button event, a tailgater at a loading dock, a behavior pattern the model has learned to flag - rather than the firehose of motion notifications that have made traditional video surveillance famously useless. Investigations that used to take hours of scrubbing tape happen by natural-language query. Compliance and evidence sharing are baked in.
Ben-Moshe likes to say the goal is for every camera to be a smart sensor. That is the line he uses in interviews. It is also the line the engineers presumably hear at every internal review.
He picked up a Technion MSc in Computer Science long before he ever sat down at Stanford GSB. The order is unusual for a Silicon Valley CEO and explains a lot about the way he talks about product.
Lumana counts McDonald's, Meta, NYU, the Minnesota Twins, and Tuff Shed among its early customers. Fast food, social media, a private university, a Major League ballclub, and a company that sells backyard sheds. The platform clearly does not care which lobby it's watching.
Lumana was founded in 2021 and did not formally emerge until April 2024. Roughly three years of quiet building before turning the marketing on. Rare in a category obsessed with launch announcements.
Headquarters in Los Gatos, R&D in Tel Aviv. Standard split for a category that draws heavily on Israeli computer vision talent, including alumni of Intel's RealSense and Mobileye.
The platform deliberately works with NDAA-compliant cameras, sidestepping the procurement and political headaches that have plagued installed surveillance for years.
No hardware lock-in. The strategic message is that the value is the brain, not the lens. The lens you already own.
"Visual AI enhances our ability to extract data and make smarter decisions."
"With Lumana, every camera becomes a smart sensor that understands, acts, and empowers people in real time."
"We're building the infrastructure for cameras to perceive, understand context and take action, not just watch."
Video analytics is forecast to grow from roughly $8 billion today to about $38 billion by 2030, per Forbes Business Insights. The growth, predictably, is being credited to AI. The less predictable part is who gets to capture it. Incumbents in the category - the camera makers, the VMS vendors - have spent the last twenty years selling boxes. Ben-Moshe is betting that the next twenty years will reward whoever sells the software that makes all those boxes useful.
If he is right, Lumana is not a security company. It is a sensor company that happens to be plugged into the cameras the world already owns. The McDonald's drive-through, the loading bay, the school hallway, the dugout. A million data points a second, tagged in real time, queried by natural language, escalated to a human only when a human is needed.
If he is wrong, he will have learned something useful while running a profitable mid-market video platform on the way. He has built quieter things before.