Cogniac Corporation

Here is a fact about the modern industrial economy that sounds made up but is not: most companies own an enormous, growing pile of images and video - from cameras on production lines, drones over rail yards, phones in the hands of inspectors - and they do almost nothing with it. The footage gets recorded, maybe reviewed if something goes wrong, and then it sits on a drive until the drive is full. It is one of the most valuable assets a company has, and it is treated like exhaust.

Cogniac Corporation, founded in Silicon Valley in 2015, is a bet that this is dumb and fixable. Its mission statement - "maximizing the value of visual data" - is the kind of phrase that could mean nothing, except that Cogniac has spent a decade making it mean something specific: point AI at the pile, and the pile starts inspecting your welds, counting your packages, reading your shipping documents, and flagging the guy on the floor who forgot his hard hat.

The interesting part is not that Cogniac does computer vision. Lots of companies do computer vision. The interesting part is who Cogniac decided should operate it. In most AI deployments, the person who understands the problem - what a good part looks like, what a bad weld looks like - is a machinist or a quality engineer, and the person who can build the model is a data scientist in a different building, or a different company, or a different country. Cogniac's product is essentially an argument that this gap is the whole problem.

So it built a no-code, enterprise-grade platform where the subject-matter expert trains the AI directly. You show the system examples, it proposes labels, you correct it, and it gets better - convolutional neural networks and automated hyperparameter tuning humming underneath, invisibly, so the machinist never has to know what a hyperparameter is. Cogniac calls the humans "the loop." The loop is the point.

Cogniac's co-founder and CEO is Bill Kish, and if the name rings a faint bell in enterprise-networking circles, that is because Kish previously co-founded Ruckus Wireless, a company that built smarter, more reliable enterprise WiFi and did well enough that he did not, strictly speaking, need to do this again. He did it anyway. His co-founder, Amy Wang, met him at Ruckus. There is a tidy logic to a networking veteran pivoting to computer vision: both are, at bottom, about moving and interpreting signals at scale, and about making complicated infrastructure disappear behind a simple interface.

The team around them is small - roughly 34 people - and deliberately mixed: engineers who build the platform, client executives who understand specific industries, and operations specialists who deploy it in the field. Cogniac describes its culture in the least fashionable way possible for a Silicon Valley AI company: people who solve real customer problems, with industry expertise, in the actual messy world. No talk of changing everything. Just watching things carefully.

The board tells you who takes them seriously. Brian Krzanich, the former CEO of Intel, sits on it. So do representatives from National Grid's venture arm and from Autotech Ventures, which has backed the company since 2017. These are not passive names; several of Cogniac's investors are strategic - corporate venture arms whose parent companies look a lot like Cogniac's ideal customers.

Cogniac's customer list reads like a tour of industries that are camera-rich, high-stakes, and famously resistant to hype: railroads, automakers, packaging plants, heavy equipment. BNSF Railway, Ford, Georgia-Pacific, Doosan Bobcat North America, and Trimac Transportation are among the named users. The Trimac deployment is a good tell for how Cogniac thinks - it is not a moonshot, it is document handling: the company's vision system reads and files the paperwork Trimac's drivers generate, turning a tedious back-office chore into an automated one.

The showpiece is automotive. Cogniac's computer vision featured in the Smart Press Shop, a "shop floor to top floor" digital-factory pilot involving Porsche, the press manufacturer Schuler, and the technology firm Syntax. When one of the most advanced sheet-metal operations in the car business wants machines that can see, a 34-person startup gets the call. That is either a strong signal or a strange one; in industrial AI, it is usually the former.

Cogniac has raised roughly $46 million. The headline round is a $20 million Series B1 that closed on October 27, 2021, led by National Grid Partners with a striking roster of strategic co-investors. The point of taking money from Cisco, Autotech, and a utility's venture arm is not just the check - it is that these backers' parent companies and portfolios are exactly the kind of camera-heavy, safety-obsessed enterprises Cogniac sells to. Strategic capital, when it works, is distribution wearing a disguise.