The company teaching security cameras to understand what they see - and act on it in real time.
Most security cameras are excellent at one thing: recording. They are far less good at understanding. A camera can capture a fall in an aisle, a fire on a loading dock, or a stranger slipping through a restricted door - and still do nothing until a person happens to scroll back through the footage, often long after it matters. Lumana, an enterprise AI company based in Los Gatos, California, was built to close that gap.
Lumana makes a video intelligence platform that layers artificial intelligence on top of an organization's existing cameras. Instead of passive recording, the system watches feeds continuously, recognizes behavior and objects, and raises alerts the moment something happens. The company describes the result as turning any camera into a real-time monitoring, detection and response system. At its core is a proprietary model the company calls VIA-1, paired with vision language models and what Lumana refers to as agentic AI - software that can perceive context, reason about it, and take action.
The scale is not small. Lumana says its platform runs analytics on more than a billion images every day, spanning behavior-based event detection, rapid investigations and operational insight. It is camera-agnostic by design, so customers keep the hardware they already own and add Lumana's intelligence on top - no rip-and-replace.
Lumana's origin traces back to Intel. Founder and CEO Sagi Ben Moshe was once among Intel's most senior Israeli executives and led its RealSense computer-vision division. When Intel wound RealSense down in 2021, Ben Moshe left and started Lumana - carrying the belief that cameras could do far more than record. He was joined by other former senior Intel computer-vision figures, giving the company deep roots in applied vision research.
Former senior Intel executive who led the RealSense computer-vision division before founding Lumana in 2021.
Former senior Intel Israel technologist, leading Lumana's engineering and platform architecture.
Computer-vision researcher and academic who guides Lumana's core AI science.
Two problems haunt traditional video surveillance. The first is time: incidents are reviewed after the fact, so the footage becomes evidence rather than prevention. The second is noise. Motion-triggered systems generate so many false alarms that security teams stop trusting them - and a team that ignores its own alerts is no safer than one with no alerts at all.
Lumana attacks both. Its self-learning models watch in real time and generate alerts based on behavior and object recognition - a person carrying a weapon, a fire, someone falling, an intruder in a restricted zone, or a worker missing required protective gear. And by leaning on a purpose-built model rather than crude motion detection, the company says VIA-1 reduces false alerts by up to 90%. The point is not just detection; it is detection people will actually respond to.
Investigations get faster too. Lumana indexes footage across more than 100 searchable attributes, so a search that once meant scrubbing through hours of video - find the person in the red jacket near the east entrance - can be answered in seconds. Teams can also configure what to watch for using plain natural language, rather than wiring up rigid rules.
What sets Lumana apart from much of the field is its camera-agnostic stance. Rivals often sell their own cameras as part of a closed system. Lumana instead treats hardware as a given and competes on the intelligence layer, which lowers the barrier for large organizations with thousands of cameras already installed across many sites.
Proprietary video intelligence model that continually learns and adapts to each environment, powering behavior-based detection and cutting false alerts.
An AI-driven video management system for monitoring thousands of cameras across sites, with a customizable dashboard that turns footage into actionable insight.
Monitor, Response, Investigate and Insight agents that raise real-time alerts, coordinate response, run investigations and surface operational trends.
Describe what to watch for in plain English - weapons, fire, falls, violence, restricted-area access - to trigger real-time alerts at scale.
Since emerging from stealth in April 2024, Lumana has signed customers across retail, healthcare, manufacturing, hospitality, education and government - including McDonald's, Meta, the Minnesota Twins Baseball Club, New York University and Tuff Shed. It competes in the fast-moving AI video security market against players like Verkada, Ambient.ai, Spot AI, Coram AI and Rhombus, differentiating on its camera-agnostic model and research-heavy team.
Lumana operates as a B2B software business. Rather than selling cameras, it licenses its AI video intelligence platform on a subscription basis to enterprises and organizations, layering the software over their existing hardware and running it through a hybrid cloud and edge architecture. Revenue is recurring and scales with deployments - a model well suited to customers running cameras across many sites. The July 2025 Series A was earmarked to accelerate go-to-market and continue product development.
| Round | Amount | Date | Lead / Investors |
|---|---|---|---|
| Series A | $40M | Jul 2025 | Wing Venture Capital (lead), Norwest, S Capital |
| Prior rounds | ~$24M | 2021–2024 | Norwest Venture Partners, S Capital |
| Total | $64M | to date | — |
Wing Venture Capital founding partner Peter Wagner joined Lumana's board with the Series A.
"Lumana transforms cameras into AI agents that perceive context, reason, and act in real-time."
— Lumana, on its video intelligence platformFormer senior Intel computer-vision leaders start the company after Intel winds down its RealSense division.
Lumana launches its AI video security platform after roughly two years of development.
Wing Venture Capital leads a $40M round with Norwest and S Capital, bringing total funding to $64M.
Signs McDonald's, Meta, the Minnesota Twins and NYU; platform reaches 1B+ images analyzed daily.
Lumana grew out of the aftermath of Intel closing its RealSense computer-vision division in 2021.
The platform processes over a billion images every single day.
Camera-agnostic design means customers keep existing hardware and add the AI on top.
You can tell Lumana what to look for in plain English instead of configuring rigid rules.
Lumana is an enterprise AI video security platform that turns existing IP cameras into real-time monitoring, detection and response systems using its VIA-1 vision model, vision language models and agentic AI.
Lumana was founded in 2021 by Sagi Ben Moshe (CEO) along with fellow former senior Intel computer-vision leaders, including CTO Ofir Mulla and Chief Scientific Officer Prof. Ron Kimmel.
Lumana has raised $64M in total, including a $40M Series A in July 2025 led by Wing Venture Capital, with participation from Norwest Venture Partners and S Capital.
Customers span retail, healthcare, manufacturing, hospitality, education and government, and include McDonald's, Meta, the Minnesota Twins, NYU and Tuff Shed.
Lumana is camera-agnostic - it adds AI on top of a customer's existing cameras - and uses a self-learning proprietary model that it says reduces false alerts by up to 90%, while enabling natural-language alerts and attribute-based footage search.