As AI expands into more industries and areas of society, a growing number of startups are building testing tools to help companies and governments assess the state of their models and identify potential compliance issues. Google Cloud chief AI scientist Andrew Moore recently predicted that the AI development pipeline will someday center on testing and validation. But Truera, previously known as AILens, says it’s building a different kind of platform, based on the research of cofounders Anupam Datta and Shayak Sen, whose explainability AI research considers causality and cooperative game theory as well as fairness and bias research.
Datta, who is on leave from his job as a professor at CMU, told VentureBeat the research that went into the creation of Truera began in 2014 when CMU researchers uncovered forms of gender bias in online advertising but lacked tools to understand how to measure that bias. The Truera platform can detect and help mitigate problems that occur in AI models, like bias imbalances, changes in data distribution that can impact predictions, or model stability over time, a feature Datta said customers are using more as a result of anomalies brought on by COVID-19.
Datta says a core tenet of Truera’s approach is that models free of heavily weighted bias toward a specific demographic group are not just more robust and capable, but also better for business.
“Even from that point, the thesis was that explainability is key to guard against social harms and to build higher-quality models. It wasn’t like either/or,” Datta, who serves as president and chief scientist, told VentureBeat in a phone interview. “In some ways, the same set of methods that let you delve deeper into models, how they are working, how … we assess or adopt them responsibly, are also the things that will help better-quality models move into production, and therefore will also be good for the world and for business. These two requirements are not in conflict. They are well-aligned.”
Today, Truera focuses on classification and regression models, as well as deep neural networks, like convolutional neural networks (CNN). The company does not currently provide services for recommendation algorithms or other specific forms of AI.
In 2019, AI startup Truera raised a $5.1 million funding round to build and grow its service. The idea was to help businesses and governments understand decisions made by AI models and perform quality assessment evaluations. Greylock Partners led the funding round, with participation from Wing VC and Conversion Capital. The funds will be used to build the first version of the Truera product, assemble the company team, and launch the platform. Initial customers are using the platform to verify regulatory compliance in highly regulated industries like banking and insurance.
Truera was founded in 2019 and is based in Redwood City, California. The company has about 15 employees.