Traverse

Traverse

Data research lab for the non-verifiable

Winter 2026ActiveB2BMachine LearningReinforcement LearningAISan Francisco, CA, USA
Company
https://www.traverse.so

Verdict

High Signal
Market Opportunity
Training data for frontier AI labs is a massive and actively funded market — Anthropic, OpenAI, Google DeepMind all spend heavily on RLHF and preference data. The specific angle of non-deterministic/ambiguous task data (law, healthcare, sales, writing) is a real unsolved problem that frontier labs are paying for. ICP is clear: frontier AI labs with budget and a need for judgment/taste training data.
Medium Signal
Founder Signal
Lance Yan was a Software Engineer at Kalshi (prediction markets startup) and Founding Engineer for Multi-Agent Infra at a stealth company, plus ML Lead at WAT.ai — real technical experience but relatively junior. Zachary Yu was a SWE at Mercor (AI hiring) and Founding Engineer at a stealth startup. Both are University of Waterloo CS students (Zachary's degree listed 2025–2030, implying he's an undergrad who left school), suggesting limited seniority. Lance's YC bio simply says 'Turning profit' which is either a flex or aspirational — no corroborating evidence in press or product data.
Low Signal
Competition
No competitor data was found in research, but the space has established players: Scale AI, Appen, Surge AI, Labelbox, and frontier labs' own internal data teams all do RLHF and training data work. The 'non-deterministic' framing is a differentiated angle, but Scale AI specifically targets this market with expert annotators. Traverse's claim to be 'pioneers' is unconvincing given the competitive landscape.
Low Signal
Product
Website is essentially a manifesto with no product visible — no demos, no pricing, no named customers, no API docs, no metrics. Just a 'Contact' CTA and vague language about being 'pioneers' in non-deterministic training data. No evidence of any shipped product or paying customers.
OverallC Tier

Traverse has a compelling thesis — training data for ambiguous, judgment-heavy tasks is a real gap — but the execution evidence is essentially zero. The website is pure manifesto with no product, no named customers, no metrics, despite claiming to be 'turning profit.' Both founders are technically credible but young (likely 20-21, dropped out of Waterloo) with limited senior experience — Kalshi and Mercor are real companies but neither founder held senior roles. The market is genuinely large and the ICP is clear, but Scale AI and others are already attacking this problem with far more resources. Without visible traction, named customers, or any product artifact, this is a high-concept early-stage bet on two smart but inexperienced founders.

Active Founders

Lance Yan
Lance Yan
Founder, CEO

Turning profit.

Zachary Yu
Zachary Yu
Founder, CTO

Solving subjectivity.

Traverse
Traverse
TierC Tier
BatchWinter 2026
Team Size2
StatusActive
LocationSan Francisco, CA, USA