Coactive AI Raises $30 Million To Make Sense Of Visual Data

Date: unknown

Location: www.forbes.com

Even for the biggest companies in the world, the huge volume of visual data they create and collect can be overwhelming to manage, whether that’s keeping track of video and photo assets for marketing campaigns or sorting through images that customers share.

To counter these challenges, companies have historically relied on human reviewers or data labeling services to categorize and analyze all of this so-called “unstructured” visual data. But by using advances in AI, a startup called Coactive AI is aiming to replace those services with software that can make sense of customers’ troves of visual content.

There’s a serious need: one of Coactive’s clients, a large retail company, would often pay for a new photoshoot to market a product it sold instead of having to sift through its huge image library to find old photos, according to Cody Coleman, Coactive’s cofounder and CEO. Now though, the startup’s software can instantly retrieve the right photo by searching for an identifying characteristic in it (say: “orange shirt”) even if it hasn’t been tagged.

On Tuesday, the San Jose-based startup announced a new $30 million Series B fundraise led by Laurene Powell Jobs’ Emerson Collective and Cherryrock Capital, a new venture fund from former TaskRabbit CEO Stacy Brown-Philpot. Other investors included Andreessen Horowitz, which led the first two funding rounds, Bessemer Venture Partners and Greycroft. The round values Coactive at $200 million, up from the $50 million mark (per PitchBook) that it received in March 2023.

Coactive’s software was built by tinkering with off-the-shelf AI models to make them adept at accurately identifying and labeling images and videos. Beyond searching through visual content, Coactive says its software also has applications in content moderation because it can accurately spot content that violates a customer’s trust and safety guidelines. Fandom, the wiki hosting service, has tapped Coactive for a content moderation tool that “eliminates time-intensive human review for every image and video” uploaded to its site, it said in a press release.

Coleman has been obsessed with computers from an early age as a refuge from a challenging childhood — he was born inside a prison and grew up in an impoverished household. He obtained bachelor’s and master’s computer science degrees from MIT and then a Ph.D. from Stanford, under the supervision of Matei Zaharia, a cofounder of $43 billion software firm Databricks. A few months before receiving the degree, he started Coactive with William Gaviria Rojas, a Ph.D. computer scientist from Northwestern and former data scientist at eBay.

Companies like Facebook, which has faced its share of scrutiny for content moderation problems, have the AI engineers to build similar tools in house. But Coactive has tapped into a revenue stream among big enterprises outside the tech sector who find it too expensive to do it themselves or pay for manual data labeling providers.

The startup would not give exact revenue figures, but said it counts dozens of customers who it estimated would boost sales more than eightfold this year. Brown-Philpot told Forbes that her firm rushed to offer a term sheet after learning that other venture capitalists had preemptively asked to invest after seeing business take off in recent months. “They were making progress, and frankly faster than most companies who were leveraging AI in some way to build their business,” she said.

Brown-Philpot, whose firm is in the middle of raising a first fund to invest in Black and Latino entrepreneurs, said her first meeting with Coleman took a detour from talking about the product to instead focus on his mission to tackle bias in AI. For example, Coactive co-created an image dataset with more than 38,000 images of items from households across a range of incomes to counter the overrepresentation of higher-income populations common among public datasets.

“If we really want to combat bias, we have to think about the data that’s going in,” Brown-Philpot said. “Coactive is not a generative AI company — they’re a [software-as-a-service] platform that is combining all the elements around AI and data to bring decisions to bear. Because of that, I think they’ve got the best potential to address the inherent bias that comes with developing any type of technology.”

Coleman said Coactive’s focus on visual data has helped it differentiate from other companies that focus more on categorizing and sorting text-based data with AI. “It’s an order of magnitude more difficult than being able to work with text data, which makes it such that a lot of people are focused on the lower-hanging fruit of text data.” The amount of data to dissect in text is akin to a lake, he said, alluding to Zaharia’s Databricks, which sells an offering called a “data lakehouse,” in which customers store and analyze their data. The scale of visual data, to continue his analogy, is more like an ocean.

“When we think about the big data tools today, it’s like having a rowboat or canoe, he said. “It’s fine to get across a lake with a rowboat, but if you told me to cross the Pacific Ocean, I’d tell you that you’re crazy and you’re going to need a bigger boat.”

MORE FROM FORBES

ForbesInside Quora's Quest For Relevance: Why CEO Adam D'Angelo Has Gone All In On AIBy Richard NievaForbesThis $2.2 Billion Startup's GPT-4 Powered AI Bot Demystifies Health InsuranceBy Katie JenningsForbes'I Want That Sweet Baby': AI-Generated Kids Draw Predators On TikTok And InstagramBy Alexandra S. LevineForbesAI-Generated Employee Handbooks Are Causing Mayhem At The Companies That Use ThemBy Rashi ShrivastavaForbesYou're Buying So Much From Temu And Shein The Air Cargo Industry Can't Keep UpBy Cyrus Farivar