How can the efficiency of AI voice agents parallel self-driving cars? Brooke Hopkins, former tech lead at Waymo and the founder of Coval, postulates that the evaluation process can, in fact, be identical.
Hopping from the realm of autonomous vehicles to AI, Hopkins identified key similarities in the challenges both industries encountered. She commented, “The struggle existed in devising innovative test practices for a new paradigm. It often felt like reinventing the wheel every time which was exactly what we had overcome in self-driving technology over the past decade.”
Capitalizing on this revelation, Hopkins launched Coval in 2024 with a mission to revolutionize performance testing for AI voice and chat agents. This innovative platform can simultaneously run thousands of simulations, testing their proficiencies in tasks such as making reservations, responding ingeniously to customer queries and more.
Coval’s technology monitors the agents through a balanced set of metrics, allowing companies to tailor the key performance parameters that matter most to them. Data gained from these evaluations can be crucial in establishing customer confidence, demonstrating that the AI agents can handle real-world situations effectively.
Hopkins traces Coval’s conception back to the Y Combinator Summer 2024 batch, and since its public launch in October 2024, the demand for their services has skyrocketed.
The San Francisco-based startup just announced a successful seed round, raising a whopping $3.3m led by MaC Venture Capital, Y Combinator, and General Catalyst. The newfound resources are earmarked for expanding the engineering team, securing product-market fit, and broadening the portfolio to include other types of AI agents in the future.
Given the increasing interest in AI agents, there is undoubtedly going to be a need for efficient evaluation systems like Coval. Hopkins believes that her longstanding experience and concrete understanding of the industry will keep Coval at the forefront.
Original source: Read the full article on TechCrunch
