Evaluation infrastructure for teams shipping ML in production.
Photo slot — deterministic SVG cover art (demo).
Modelyn builds evaluation infrastructure for teams running machine learning in production. Instead of a dashboard of proxy metrics, it gives teams versioned eval suites that run on every model change — the same discipline software engineers expect from CI, applied to models.
Based in Berlin since 2021, the company grew out of consulting work for industrial ML teams and kept the same sensibility: evals are written against real failure cases, not benchmarks. Customers use Modelyn to answer one question with confidence — did this change make the model better, or just different?