Just went through Anduril's DS loop for a data scientist role on their autonomy analytics team. The loop is not what I expected, partly because Anduril's data science function is closer to applied research than business intelligence.
Loop structure: Recruiter screen (30 min, standard motivation + background) Technical phone screen with a DS (45 min, SQL + stats) Onsite: 4 rounds over a day (SQL deep dive, case study, ML/stats, behavioral)
SQL round:
Harder than average. They gave me a schema representing event logs from an autonomous system. Window functions, CTEs, and aggregation were all in play. One question had a subtle gotcha around NULLs in a join that I almost missed. They're testing whether you can actually use SQL to analyze real messy data, not whether you know the syntax. Practice writing queries against normalized schemas with missing data, not toy examples.
Case study:
This was the most Anduril-specific part. The scenario: a sensor fusion algorithm is producing more false positives than expected in certain environmental conditions. How would you investigate this, what data would you pull, what analysis would you run, and how would you communicate findings to engineers? I think they're really evaluating whether you can bridge between data and engineering and whether your analysis would actually drive a decision.
Stats/ML round:
Fairly standard: A/B testing (what's your minimum detectable effect, how do you handle interference between units), bias-variance tradeoff, discuss a model you've built end-to-end. No whiteboard coding of ML algorithms but be ready to explain what's happening under the hood of gradient boosting if they ask.
Comp I saw: A friend who joined as DS2 in early 2026 in the Costa Mesa office said their total comp was around $180-195K with a meaningful equity component. That's not FAANG-level but it's strong for defense tech, and the mission draw is real for some people.