Brooke Hopkins

About Brooke Hopkins

Brooke Hopkins is the founder of Coval, a San Francisco company that builds evaluation, simulation and monitoring tools for voice AI agents. Before Coval she led the evaluation infrastructure team at Waymo, which built the developer tools for launching and running self-driving simulations. She argues that autonomous voice agents and autonomous vehicles share the same problem: getting reliably from point A to point B across every path in between, with many models working together.

Coval launched in 2024 and raised a $28 million Series A in June 2026. Its customers range from AI companies such as Perplexity to Fortune 500 enterprises, including Upstart, which uses Coval to review and improve its loan verification voice agent.

Career Highlights

• Founder of Coval, evaluation and simulation for voice agents
• Led evaluation infrastructure at Waymo
• Raised a $28 million Series A for Coval in June 2026

Notable Positions

On Evals as Product Definition

Hopkins says teams should spend most of their eval time deciding what the agent should and shouldn't do. She groups metrics into three questions: did it reach the goal, did it follow the required steps, and how good was the conversation. Metrics should come from what human reviewers actually care about, not generic measures like word error rate.

On Shifting Left With Simulation

Coval re-runs failures as voice-to-voice simulations with different accents, background noise and interruptions, so fixes are proven before customers hear them. She says enterprises go wrong when their release gates make it easy to launch once but hard to change a model or prompt later.

On Self-Healing Agents

Her end state is a loop where agents propose fixes from eval data and fine-tuning improves the system over time, with every change still gated by human review.

Key Quotes

• "Evals are really the product definition." (on where teams should spend their time)

Related Reading

• Upstart and Coval: Voice AI Evals in Regulated Lending - Full breakdown of the talk
• Agent Evaluation - How organizations test agents before and after launch
• Human in the Loop - Keeping people in control of automated changes