Sepal AI

Sepal AI was a San Francisco AI data company founded in 2024 by Robi Lin, Kat Hu and Fedor Paretsky. It built expert-created training data, model evaluations, safety tests and reinforcement-learning environments for advanced AI systems. Mercor acquired Sepal AI in February 2026, including its team, technology and expert network.

Sepal no longer operates as a separate contributor platform. As of July 2026, its former main website no longer serves an operating company site, and current contributor opportunities belong under Mercor. 

AI data training work at Sepal AI

Before the acquisition, Sepal focused on projects where specialist knowledge mattered. Its network included people from science, engineering, medicine, finance, business, law and other professional fields. Tasks could involve writing difficult questions and ideal answers, evaluating model reasoning, checking facts, designing benchmarks, red-teaming unsafe behavior and reviewing another expert’s work.

Sepal’s Y Combinator profile described work across training data, evaluation datasets, safety and reinforcement learning. This was generally more demanding than simple labeling: projects could require advanced education, professional experience or clear evidence that a contributor could explain a specialized judgment.

That history remains relevant because it helps explain the kinds of expertise and project methods Mercor gained through the acquisition. It does not mean a former Sepal role or expert-hub listing is still accepting applications.

What is it like to work at Sepal AI?

Sepal recruited experts for individual opportunities rather than maintaining a universal queue of microtasks. The arrangement, location, workload and schedule depended on the project. Rates could be hourly or output-based, and acceptance into the expert network did not guarantee immediate or continuous assignments. Availability could change with client demand and specialist fit.

Recent developments and media coverage

The acquisition announcement confirms that Mercor acquired Sepal. Any current offer connected to the former network is therefore expected to identify Mercor or another clearly named contracting entity, along with the relevant onboarding and payment process. Old Sepal branding may still appear in historical messages or links during the transition.

Member ratings and reviews now belong with Mercor so that current experiences are not split across an inactive company page and the platform that continues the work. Historical Sepal experiences can still provide context when a review makes the timing and company relationship clear.

Application process at Sepal AI

Sepal’s earlier applications could request contact details, a résumé, education, professional profiles and evidence of subject expertise. Depending on the project, screening could add a call, identity or credential verification and a written assessment involving a subject problem, AI-response evaluation or scoring explanation.

For current vacancies, application details and Thriveth member experiences, continue to the Mercor company profile.

Sources: Company website · Acquisition announcement

Information last updated August 2026.