TruEra, a company that provides an AI quality management system for optimizing, explaining, and monitoring machine learning models, has secured $25 million in a Series B investment headed by Menlo Ventures. Existing investors Greylock Partners, Wing Venture Capital (which led a $12 million Series A round in late 2020), Harpoon Ventures, Conversion Capital, the Data Community Fund, and new investors Forgepoint Capital and the B Capital Group’s Ascent Fund, as well as new investors Forgepoint Capital and the B Capital Group’s Ascent Fund, participated in this round. TruEra has already raised $42.3 million in total.
“We feel the next great problem in AI is the quality challenge,” said Will Uppington, CEO and co-founder of TruEra. “AI is at a crossroads: there are many potential but also many hurdles in making AI function in the workplace.” And we believe that this is the primary impediment to AI systems being used in the real world and delivering on KPIs.”
Uppington claims that it’s not only difficult to design and build high-quality models in the first place, but there are also ongoing concerns about trust, transparency, and fairness when it comes to putting models into production — and rising regulatory pressure on AI fairness is causing businesses to pause because they need to put auditable systems in place to comply with these rules. And once a model is ready for production, organizations must verify that the quality of the data remains consistent, even if part of the underlying data changes.
According to TruEra, an enterprise AI quality management system should tackle these issues straight on, beginning with tools that developers can use when training models so that they can test and review them long before they get into production. To do this, the company’s service, for example, may be linked directly into the Jupyter notebooks that most data scientists currently use to develop their models. Uppington explained, “We’re in the place where software development was in the 1990s, before you had tools and agile development processes.”
“Data science is a waterfall process, and the models are still a bit of a black box.” This lowers the quality of your development process, just like it did in the 1990s with software development. […] With this type of thorough, quick quality testing and making it incredibly easy for data scientists to use, we believe we can assist the world get to those better tools and more agile-like development.” Since raising its Series a round in late 2020, the firm claims that sales has increased by more than 5 times. This is due, in part, to the fact that many businesses are now at the stage in their AI journey when they want to put models into production and are encountering quality issues, according to Uppington.
It’s an excellent moment to be in the AI quality field, especially given the regulatory climate and several high-profile failures (think Zillow). TruEra’s co-founder and Chief Scientist Anupam Datta has done some of the early academic work on AI explainability as a professor at Carnegie Mellon, according to Menlo Ventures partner Tim Tully, the former CTO of Splunk.
“What would I use if I were a machine learning data scientist?” “I looked at all of the firms and was searching for one that gave depth and tackled the problem through the lens of the model, rather than the other way around, which I believe is utterly incorrect,” Tully said. “And, to be honest, I think a lot of the items are only scratching the surface of the problem.” I’d want to go into it further, and I’d like to see some exclusive research that sets us apart from the competition.” Tully noted that using open source Python modules and slapping a user interface on top of them won’t cut it in this industry.