Hey everyone! I'm new to the MLOps side of things and currently helping my team evaluate monitoring solutions. We're at a Fortune 500 retailer, and we need to scale monitoring for dozens of models (recommendations, demand forecasting, fraud detection).
I've heard great things about Arize, but some senior engineers are suggesting we just build custom dashboards with Grafana/Prometheus. Since I'm still learning, could someone break down the real trade-offs here?
For example, is Arize mostly about the "out-of-the-box" features for drift, data quality, and performance? Would a custom stack be better just because we have very specific, existing pipelines? I worry about building something that becomes a maintenance nightmare.
Thanks for any guidance! 😊