platoseedplatoseed
Research lab solving non-verifiable work
Traverse is a research lab that partners with frontier AI labs to produce training data for non-deterministic, long-horizon tasks. Its focus is on capturing real expert reasoning in real environments to teach models judgment, taste, and context for economically valuable work.
Traverse acquires and generates reinforcement learning environments and training data that document and preserve expert decision-making processes in ambiguous, real-world contexts. They emphasize capturing reasoning, constraints, and contextual signals from actual experts operating in real environments to create training signals that teach models how judgment and long-horizon reasoning work in non-deterministic tasks.
Who it’s for: frontier AI labs / research teams developing models requiring non-deterministic, judgment-focused capabilities
Hiring/traction mention via Careers; partnerships with frontier AI labs implied
We give AI labs the training data that enables their frontier models to acquire taste
Traverse builds data factories that capture real human judgment in high-fidelity, real-life work environments to train frontier models on ambiguous, taste-dependent tasks (law, healthcare, sales, writing, etc.). It focuses on turning subjective reasoning and long-horizon decision making into training data at scale for frontier-model training.
Formerly “Clice AI”, “Clice”
AI driven professional engineering firm
The data infrastructure for real estate
