Judgment Labs raises $32M

Judgment Labs said on May 12 that it had raised $32 million in combined seed and Series A funding to build software that helps AI companies improve agents using production data. Lightspeed Venture Partners led both rounds, the San Francisco startup said, with Nova Global, SV Angel, Valor Equity Partners and Dynamic also participating. (morningstar.com) The company said the product is already in use at agent-native companies and that the new capital will go toward its research team and platform. ### Why did this company raise two rounds at once? May 12 was the date Judgment Labs disclosed the financing as a combined seed and Series A round rather than as separate announcements. Lightspeed led both financings less than six months apart, according to the company’s announcement, a detail that shows the investor returned quickly with more capital. (morningstar.com) James Alcorn, a partner at Lightspeed Venture Partners, said the firm backed the company twice because of what he called “extraordinary” results. Alcorn said Judgment was addressing “the hardest problem in the agent stack” — measuring and improving systems that plan, use tools and retain memory. (morningstar.com) ### What does Judgment Labs actually sell? Judgment Labs’ documentation says its product traces agent inputs, outputs, tool calls and model calls, then uses agent-judge evaluation to detect failures and validate fixes against production cases before release. The company maintains an open-source SDK called Judgeval and offers a platform that customers can access through API keys after requesting access. Paraform’s company profile for Judgment Labs describes the category as “Agent Behavior Monitoring,” or ABM. (morningstar.com) The profile says the software is meant to surface behavioral anomalies such as instruction drift and context-retrieval loss in production, rather than only logging latency or exceptions. ### Why are AI agent companies looking at production data now? Judgment Labs said the shift from chatbots to more autonomous “deep agents” is changing how developers measure quality. (morningstar.com) In its funding announcement, the company pointed to systems such as Anthropic’s Claude Code, OpenAI’s Codex and Cognition’s Devin as examples of agents that handle longer, multi-step tasks, including coding, web browsing and follow-up questions. The company said older evaluation methods were built around a single prompt and a single answer, while newer agents generate long reasoning traces, tool use and memory over time. (docs.judgmentlabs.ai) That is the gap Judgment says it is trying to address by turning real-world telemetry and outcomes into labels, monitoring signals and model improvements. (paraform.com) ### Who founded Judgment Labs? Alex Shan, Andrew Li and Joseph Camyre founded Judgment Labs in 2025, according to company profiles and coverage of the financing. Shan is chief executive, Li is chief scientist and Camyre is chief technology officer. The company has said the three founders were childhood friends. Paraform lists the company at 19 employees in San Francisco and says it has raised more than $30 million across two rounds in the past five months. (morningstar.com) The investor list on that profile includes Lightspeed, SV Angel, Valor Equity Partners, Nova Global, Chris Manning, Michael Ovitz, Michael Abbott, Cory Levy and Kevin Hartz. ### What should readers watch next? Judgment Labs’ docs page says the company is offering access to its platform through demos and API credentials rather than broad self-serve signup. (morningstar.com) The same materials show the company is pushing developers toward tracing, evaluation and validation workflows tied directly to live agent deployments. The next concrete step is hiring. Paraform lists open roles including forward deployed engineer, research engineer, senior cloud infrastructure engineer and agent product engineer, all based around the company’s San Francisco operation. (finsmes.com) (docs.judgmentlabs.ai) (paraform.com)