More companies are turning to Artificial Intelligence (AI) and Machine Learning (ML) to become more data-driven and agile in their operations. The adoption has accelerated at such a pace, Gartner predicts that by 2024, three-quarters of organizations will operationalize AI in some form.
Historically, customers have opted for off-the-shelf ML applications. Yet, these “point” ML solutions often fail to meet enterprise companies’ ever-growing complex and specific use case requirements.
As a result, customers are now demanding custom ML solutions that are cost-effective, easy to iterate, and capable of turbo-charging the domain knowledge of their employees.
Enter Cogniac, a visual intelligence platform that applies the latest AI models, cloud computing, and big data management to provide visual operations intelligence capabilities. Cogniac’s platform helps customers automate their end-to-end ML pipeline with repeatable, consistent workflows enabling faster experimentation, development, and deployment, whether dealing with X-rays, security camera images, documents, or miles of railroad tracks.
Sharing a vision to make infrastructure and business processes more efficient through digital transformation, Cisco is excited to announce our investment in Cogniac’s latest financing round. To talk more about the company’s solution and business philosophy, I recently sat down with Cogniac CEO Chuck Myers.
From wireless to neural networks
Cogniac Co-founders Bill Kish and Amy Wang met at Ruckus Networks, an early enterprise wireless company Kish co-founded to build smarter, more reliable WiFi.
After spending 11 years leading Ruckus, Kish was looking to identify his next opportunity. Myers explains, “The core premise of Ruckus was to build an enterprise wireless network that anybody could install in a building, a convention center, a hotel or an office without needing radio frequency experts. The genesis was AI and ML built into the routing technology. Then around 2016, Bill said, ‘What’s another technology that’s similar to radiofrequency with the same kind of problems that RF had back in the 1990s and 2000s?”
Kish found the answer in machine vision. The decades-old vision sector was dominated by a few legacy companies who were still using camera-based solutions focused on pixel counting and pixel masking to perform simple quality-check tasks with a go/no-go gauge. Their solutions required expensive light fixtures and programming by experts and suffered from low tolerance for any image or calibration variation. Accuracy declined over time.
Kish began to explore machine vision as a more dynamic solution to scan a broader range of images and video data while requiring no or low programming and model training. Convolutional neural networks offered promise.
“Whereas the traditional machine vision systems are a mask, literally counting pixels, artificial intelligence considers billions of calculations of how each pixel relates to the next. As you start to cull it down closer to what you’re looking for, it becomes like a human and starts to recognize issues on its own,” says Myers.
Empowering less technical users to deploy ML models
Catering to a clientele ranging from automotive and rail to security and government, Cogniac’s no-code, visual intelligence platform helps its users automate the visual inspection and analytics tasks of peering over thousands or even millions of images and video data to identify defects in the product.
The platform packages next-gen deep learning and hyperparameter optimization in an easy-to-use interface that supports cloud, on-premises, and hybrid deployments with enterprise-grade management capabilities.
Users rave about the platform’s key features, such as using fewer training images, speeding up ML model creation, allowing domain experts to provide input, and offering a feedback loop to train and improve the ML models continuously.
Much of the work centers around annotating an image with AI. “With very few images of a known part, if you’re in the automotive space, for instance, the system will start to identify material splits and feed that information back to the business analyst. The more this occurs, the better the system gets. So, you don’t have to hire a team of 500 people sitting in a dark room annotating images,” Myers says.
Targeting visual inspection to deliver ROI
Visual inspection of assets and processes is a ubiquitous use case that has driven a huge demand for image processing. For instance, one of Cogniac’s early customers, a major railroad, must review 30 million images of train wheels a month, scanning for wheel cracks to avert a derailment.
“In manufacturing inspection, people are great inspectors for about the first three minutes in which they’re probably accurate 80% of the time,” Myers says. “But people get distracted and bored, and they start to miss. However, AI technology doesn’t go to sleep, and it retains all the information. Every time our system finds one of those errors, the information gets built back into the model.”
Best of all, the platform is scalable and easy to administer. With minimal training, users can deploy the system on their own and build their own applications without having to hire a team to implement.
Scaling the company for the future
After introducing the enterprise-grade platform across a variety of verticals, Cogniac brought Myers onboard in 2020 to lead and scale the company. Having led several growth-stage private and public companies as founder and CEO, Myers’ early leadership training came from an unconventional place, spending his teen years sailing boats professionally with global business leaders who became “unbelievable mentors.”
Having led a defense and intelligence contractor, one of the original telematics companies, and a couple of software startups, and turned around and exited two other companies, Myers was not looking to join an early-stage startup at the time he was recruited.
However, that changed when he met Kish and discovered Cogniac. “I was mesmerized by the technology. Here was a company that had raised just $10 million funding. They had some of the best AI technology and yet no sales or marketing. Nobody had heard about them. And, I spent a lot of time with Bill and looked at his track record. I thought I've had a lot of good singles and doubles. He's had three home runs.”
Early in the market uptake process within an industry that was also developing, Cogniac had an enterprise-class product with not enough name recognition and very little money invested.
Myers plans on changing that. Over the last two years, he’s raised more investment, brokered partnerships with SAP and Nvidia, and added several new customers.
Going forward, Myers and Kish want to continue embedding industry-leading AI and edge capabilities in their solution and nurture an ecosystem of system integrators and channel partners to augment their direct selling efforts. The company will ramp up hiring and strive to build the industry’s leading AI team. Myers’ motto is to “hire good people and feel good about the people you hire and help them succeed in the job that you think they're going to be best at.”
Cisco is excited to join an impressive group of investors as, together, we support Cogniac in its growth journey.