Seth Kimmel, founder of Sutro, has open-sourced jev-align, an experimental command-line tool that calibrates the Jev foundation model to a team’s decision criteria.
The tool uses accumulated human labels and optional rationales to run GEPA, then proposes improved function definitions. It requires Python 3.11 or newer and is distributed under the Apache-2.0 license.
Jev is TypeSafe AI’s first “System One Model,” designed to produce structured, typed probabilistic outputs rather than conventional text generation. The jev-align repository says Sutro is not affiliated with TypeSafe AI.
Software-verified metadata for the announcement post records 93,152 views, 733 likes, and 61 reposts at the time of candidate verification.
Sutro describes its platform as a way to build and deploy AI Functions at scale, align models to human expert judgment and select a cost-effective production model. Its site says users can upload an unlabeled dataset, define a task and add feedback on difficult cases.
According to Tracxn, Sutro was founded in 2024, is headquartered in San Francisco and has not raised external funding rounds. Tracxn ranks it 71st among 206 active competitors in its category and lists Atlan, Voyage AI and Monte Carlo among its top competitors.
TypeSafe AI, Jev’s maker, emerged from stealth on September 15, 2026 with $40 million in seed funding led by DCVC. GEPA is also available as a separate general-purpose open-source package for optimizing prompts, code and other text artifacts.
Kimmel announced the news in a post on X.
