A basic HR chat, then a tech screen which was mostly Python data manipulation and SQL window functions. The final loop is four rounds and it is pretty exhausting. The main filter is the panel presentation. You get a fake customer scenario a few days prior, and I had to pitch a lakehouse migration. They will literally interrupt you mid-sentence, acting like a skeptical CTO or a defensive data engineer, to see how you handle pushbacks.

After that is an architecture deep dive. You cannot fake this part. They ask deep Spark internals like managing data skew, broadcast joins, and how the Delta Lake transaction log actually works. There is also a live ETL coding round and a behavioral interview mostly about handling core tech and sales teams.

To prep, I actually binged a bunch of Databricks summit videos on YouTube to learn their specific jargon. The panel round is impossible to practice alone, though, imo. I searched everywhere for a real SA to mock with and finally got one on Prepfully. The guy was super helpful; he roleplayed the skeptical customer perfectly and told me my architecture pitch was way too technical for a CTO persona, haha. Definitely do a dry run, I would say, if you can, because that presentation round will make or break the offer. Or at least that is what I think.