Most companies struggle to extract value from their data. Several years ago, Forrester reported that between 60% and 73% of the data belonging to the average business goes unused for analytics. This is because data is siled or otherwise pigeonholed due to technical and security considerations, making it difficult—if not impossible—to apply analytical tools.
Anna Pojaves and Tyler Marin, engineers who previously worked at Y Combinator-backed startups HiTouch (a data syncing platform) and FairSquare (a health insurance tool) to discover Later I was inspired to try my hand at solving the data value problem. Many companies were “locked out” of analytics strategies due to engineering constraints.
“We've found that a significant portion of the market, particularly in regulated industries like healthcare and finance,” are struggling with data analytics, Marin told TechCrunch. “The vast majority of corporate data today doesn't fit into a database; it's sales calls, documents, Slack messages and so on. And, given the scale of these companies, off-the-shelf data models are usually not enough. “
So Pojawis and Maran founded OmniAI, a set of tools that transform unstructured enterprise data into something that data analytics apps and AI can understand.
OmniAI integrates with a company's data storage services and databases (e.g., Snowflake, MongoDB, etc.), generates data in-house and allows companies to run the model of their choice — for example On, a large language model — on data. According to Marin, OmniAI performs all of its operations in the company's cloud, OmniAI's private cloud or on-premises environments, which apparently provide better security.
“We believe that large language models will become essential to a company's infrastructure over the next decade, and holding everything in one place only makes sense,” said Marin.
Out of the box, OmniAI offers integration with models including Meta's Llama 3, Anthropic's Claude, Mistral's Mistral Large and Amazon's AWS Titan to automatically correct data-sensitive information and general AI-powered applications. Making cations. Customers sign a software-as-a-service contract with OmniAI to enable the management of models on their infrastructure.
It's early days. But Omni, which recently closed a $3.2 million seed round led by FundersClub at a $30 million valuation, claims it already has 10 customers, including Klaviyo and Carrefour. Annual recurring revenue is on track to reach $1 million by 2025, Marn said.
“We're an incredibly lean team in a fast-growing industry,” Maren said. “Our bet is that, over time, companies will choose to run the model alongside their existing infrastructure, and Model providers will focus more on licensing model weights to existing cloud providers.”