I have a Flink Job with keyBy processing functions. Problem is that it's unevenly distributed. Unfortunately, I can't know for sure which key will have more data. Is there a way to make a better distribution between slots?

Salting/bucketing can be used if you don’t have any control over the key. Check this blog: https://eng.lyft.com/gotchas-of-stream-processing-data-skewness-cfba58eb45d4