«Test ML? That's a bit too broad»: how I built a QA framework for ready-made m…

«Test ML? That's a bit too broad» — this is how the author begins her story about building a QA framework for ready-made models. Six months ago, she had already created her own agent for checking technical specifications, and now she took on a new task. In the post, she explains how she approached testing already trained models and what came of it. This is practical experience that can be applied to your own projects if you need to systematically check the quality of ML models. Conclusion: even a too broadly defined task can turn into a working tool.