Curious if anyone has any experience with Flyte and how it compares to Metaflow? We chose Metaflow because of its human centric approach to doing Data Science and because its more intuitive to adopt and the documentation was much cleaner. But Flyte also has a bunch of great plugins and operators much like Airflow.

I’d be curious to hear if folks have experiences too. When it comes to plugins and integrations, our approach in most cases is to document best practices rather than creating layers of abstractions, unless there’s a compelling reason to do so

I tried using Flyte before Metaflow and found it was pretty difficult to get even a basic example working due to the poor documentation back then(not sure if things have changed since then).

I think Metaflow is much better at building automated ML training and validation workflows (experiment tracking + workflow orchestration 2in1). Flyte doesn't have as nice of a UX, but it is better suited for building general pythonic event-driven chainable workflows.

I tried Flyte before. The setup was definitely not trivial, even using the automated setup with Opta. I felt that it required too much infrastructure to actually work (scheduler, EKS, RDS etc), which would lead to operational pain or lack of support if there were issues. With Metaflow I mainly have to deal with Batch (and maybe EventBridge, but still AWS offerings).