Databricks · Primly Community

Databricks data scientist interview (SQL + case + stats), what the loop looks like

ds_dmitri · 5 replies

Just wrapped the Databricks DS interview process for a senior data scientist role on the product analytics side. Here's the breakdown for people searching for recent data points -- most threads I found were 2-3 years old.

The loop was 4 rounds after the recruiter screen.

SQL / technical coding round (60 min). Two SQL problems, both on a shared editor. First was a medium-complexity aggregation: multiple joins, window functions, filtering on derived columns. I'd call it LeetCode medium SQL. Second was harder: a self-join plus LAG/LEAD to track sequential events in a user funnel. Took me most of the remaining time. They're not testing whether you know Databricks SQL syntax specifically (it's similar to Spark SQL), but they want to see you're fluent at analytical SQL.

Also a quick Python section: one data manipulation question using pandas. Write a function to compute a rolling 7-day average grouped by user. Nothing tricky, just fluency.

Statistics and probability round (45 min). Heavier than I expected. A/B testing: what sample size do you need? They walked me through a scenario with low base rates and asked about power calculations. Also asked about p-value misinterpretation (classic) and a question about Bayesian vs frequentist approaches to a product metric. I work in frequentist land day to day so I fumbled the Bayesian question a bit but it didn't seem to tank me.

Case study / analytics thinking round (60 min). Given a scenario: "Databricks usage among a cohort of enterprise customers dropped 15% month-over-month. Walk me through how you'd investigate." Very standard DS case. I went: data quality check first, segment by customer size and product area, check for external factors (did we change pricing? A feature break?), identify leading vs lagging indicators. They pushed on each step.

Behavioral (45 min). Same format as the eng behavioral. Real person from the DS org, ownership and impact questions. I had one story about a time my analysis surfaced a finding that a major product team didn't want to hear. That went over well.

Comp target they shared during the process: roughly $200-240k total comp for senior DS in SF, heavy on RSUs. Felt competitive for the market in 2026.

5 replies

analyst_ana

The funnel tracking SQL question with LAG/LEAD is everywhere right now. That's basically a standard for any analytics-heavy DS interview. If anyone is prepping: be able to write a query that finds users who hit step A then step B within N days without hitting step C. Variations of that show up constantly.

content_cole

$200-240k total comp for senior DS in SF in 2026 sounds about right. I've seen DS offers at Databricks come in around there for mid-senior. For comparison, a friend's offer (also senior DS, ML-leaning) was ~$230k with ~40% in RSUs on a 4-year vest, no cliff.

ds_dmitri

Yeah the RSU split felt high to me relative to base but the vesting structure was standard. Worth modeling out at the current pre-IPO valuation if you care about that. It's not a public stock so you're taking some variance there.

returner_ren

Was the stats round completely verbal or did they have you write anything out? I've been prepping stats concepts but get nervous doing power calculations in my head out loud.

ds_dmitri

Mix of verbal and scratchpad. For the power calculation they let me write out the formula steps. I don't think they expected mental math, more that you know the variables (effect size, alpha, power) and can reason about the direction of the tradeoffs.