There is a genre of technology company that likes to tell you it is transforming an industry, and there is a smaller, more interesting genre that just tells you a number. Vivity AI is in the second genre. The number is roughly two million dollars, which is about what a single unplanned shutdown can cost a petrochemical plant, and the entire pitch of the company is that it would like to help you not lose that money. This is refreshing, because "we will save you $2 million per shutdown you don't have" is a claim a plant manager can actually evaluate, unlike "digital transformation," which is a claim that mostly evaluates the person saying it.
Vivity AI was founded in 2022 and does something that sounds simple and is not: it runs artificial intelligence on the factory floor. Not in a cloud, not in a data center in another state, but on-site, next to the machines. Its platform, called Vivity Edge, is described as a containerized, modular system that talks to the standard equipment already bolted into industrial plants - the programmable logic controllers (PLCs), the industrial IoT sensors, the CCTV cameras. The technical translation of that is "we made the AI meet the machines where they already live." The business translation is "we did the annoying integration work so the customer doesn't rip anything out."
This distinction matters more than it sounds. A lot of AI works beautifully in a demo and then encounters a real factory, which is full of hardware that predates smartphones and was never designed to have opinions about neural networks. The interesting engineering at Vivity is less the model and more the plumbing - getting inference to happen reliably, on cheap edge hardware, next to a PLC that has been running the same loop since before the founders finished school.
"Zero accidents, zero regulatory violations." That is not a slogan. It is a safety director's job description.
What It Actually Does
The company sells a handful of modules that all sit on the Edge platform. Vivity Safety AI watches for accidents in real time and is aimed, in the company's words, at "zero accidents, zero regulatory violations." Vivity Analytics reads the operational data and looks for the equipment failure that is about to become expensive. And a system called Dynamic Visual Intelligence - DVI - sends drones over shipyards and uses computer vision to figure out where each enormous ship block is and how far along the assembly has gotten.
If you want to know why any of this is hard, consider the false-alarm rate. Vivity says its industrial deployments run below 0.1%. In a consumer app, a false positive is mildly annoying; you dismiss a notification and move on. In a refinery, a system that cries wolf gets ignored, and a system that gets ignored is worse than no system at all, because it was supposed to be the thing that caught the alarm that actually mattered. So the entire value of a safety product collapses into a single unglamorous statistic: how often is it wrong when nothing is happening. Vivity has decided, correctly, that this number is the product.
The vision models come from Ultralytics, whose YOLO family Vivity uses to hit what it reports as 99.8% object-detection accuracy at 60 frames per second - fast enough to matter in an environment where a half-second of latency can be the difference between catching something and reading about it later. The cameras themselves are, in some deployments, explosion-proof, because the places Vivity operates are the kind of places where a spark is not a metaphor.
The Money, and Who Believes in It
In January 2023, Vivity raised roughly $8 million. The notable name on the cap table is Hanwha Systems, part of one of South Korea's largest industrial groups - the sort of strategic investor that brings not just money but customers, factories, and a very direct opinion about whether the product works. Plug and Play, the accelerator, is also in. For a company selling to heavy industry, a heavy-industry backer is close to the ideal shareholder: they can pick up the phone and get you into a plant.
The founder and chief executive, Yu-Sung Chang, arrived at this from an unusual direction. Before Vivity he held technology leadership in Korean e-commerce, a world of recommendation engines and shopping carts. He could have built another consumer app. Instead he pointed AI at shipyards and refineries, which is either a strange career choice or a shrewd one, depending on how you feel about competing in crowded markets versus enormous, unglamorous, underserved ones. Vivity is a bet on the second.
The company is small - about 25 people - and split across two continents, with an office in the San Francisco Bay Area and another in Seoul. In 2024, Frost & Sullivan named it Entrepreneurial Company of the Year in industrial AI, which is the kind of recognition that signals a company has moved past the demo stage and into the part where it has to actually deploy and get paid. The claimed savings pile up from there: about $2 million per avoided shutdown, another $300,000 in productivity once a plant goes fully online, and, in shipbuilding, something north of $5 million a year from letting drones and models do the counting that people used to do by walking around.
None of this makes Vivity a sure thing. Industrial sales cycles are long, incumbents are sticky, and "our AI prevented the disaster that therefore did not happen" is a famously awkward thing to put on an invoice. But the company has picked a real problem, attached itself to a real number, and found a backer who owns the factories. In a field full of AI companies explaining how they will eventually be useful, that is a comparatively grown-up place to stand.
