🗼
Trajectory List
Explore roles at AI companies on breakout trajectories. Updated daily.

Judgment Labs

Judgment Labs builds the infrastructure layer that makes AI agents get better over time in production. Their platform ingests everything a deployed agent does — tool calls, reasoning traces, memory reads, retries, and outcomes — and turns that raw experience into structured signals: recurring failure modes, behavioral rubrics, and auto-generated evals. Teams can then identify exactly where agents go wrong, validate fixes before deploying, and close the loop by feeding that environment data back into post-training (RL and SFT). Unlike legacy observability tools built for single-turn chatbots, Judgment evaluates entire multi-step agent trajectories, not just final outputs. The open-source framework (Judgeval, Apache-2.0) drives developer adoption; the hosted cloud platform adds AutoRubrics, dashboards, and enterprise features. Primary buyers are AI-native companies deploying agents in high-stakes workflows — finance, legal, and operations. The company was founded in 2025 by CEO Alex Shan, Chief Scientist Andrew Li, and CTO Joseph Camyre — all under 24 — with deep roots in Stanford NLP research, TogetherAI, and Datadog infrastructure.

22 employees · 5 open roles

Website · Careers

Open roles