went through the GitLab DS loop for a mid-senior DS role on the product analytics team earlier this year. not a lot of specific info out there so here's what i found.
the structure: recruiter screen (30 min) hiring manager (45 min, background + how you think about problems) SQL + technical interview (60 min, live) product case interview (60 min) statistics / ML concepts interview (45 min) behavioral / values (45 min)
five substantive rounds is a lot. budget for a full day if you're doing it back-to-back.
SQL round: this was harder than average. not 'SELECT * FROM users WHERE date > X' stuff. we did: a multi-table join across user events, pipeline runs, and project metadata to answer a business question about activation drop-off. then a follow-up where i had to rewrite a slow query (it had an N+1 pattern baked in). then a window function question: 'for each user, find the time between their first pipeline run and their first successful pipeline run.' that last one is where i nearly fumbled.
the SQL environment was just a shared doc. no live query runner. so you're writing it out and explaining, not running it.
product case: prompt: 'GitLab's free-to-paid conversion rate dropped 8% last quarter. walk me through how you'd diagnose it.' totally open ended. they want funnel thinking, cohort thinking, and hypothesis generation. i ran through it for 25 minutes and then they started asking 'what if i told you the drop happened only on mobile?' type steering questions. classic case interview cadence.
stats / ML interview: no coding. pure concepts. questions: explain precision vs recall in the context of a spam classifier. explain how you'd detect if an A/B test had a novelty effect. 'your model shows 92% accuracy, why might that number be misleading?' (imbalanced classes, obviously, but they want you to reason through it). also: bias-variance tradeoff, when would you choose a simpler model over a complex one.
overall difficulty: harder than i expected, especially the SQL. prep with window functions and multi-table aggregations. the case and stats rounds were well within normal DS interview difficulty.
comp: my offer was around $165k base for mid-senior level, US remote, in 2026. RSUs on top. not FAANG numbers but fully remote and the work is interesting if you care about devtools metrics.