Just finished. Did not get the offer, but I learned a lot, so sharing.
Rounds in order: Recruiter screen, 20 min, mostly logistics and the "why Palantir" question Technical phone screen, 45 min, two medium algo problems (one graph, one interval merging) Onsite: 4 rounds back to back over two days Algo (harder, a variant of topological sort with a twist I hadn't seen) System design (design an alerting system for anomaly detection across distributed sensor data) Decomp (the unique one, below) Behavioral / fit, 45 min
The Decomp round was with a senior engineer who gave me a scenario about tracking movement patterns across a city for public health. No code. Just: break it down, what data do you need, what are the edge cases, how do you handle missing data, what are the tradeoffs between a batch and streaming approach. We went back and forth for 50 minutes. It felt like pairing with a colleague, which was actually kind of fun.
What killed me was probably the behavioral round. They really probe on the defense/government work angle. "Are you comfortable with Palantir's work in [specific area]?" Be prepared to have an actual opinion. I hedged and I think it cost me.
Comp for L3 in NYC: my recruiter quoted a base range of $155k-$185k with significant stock, but I never got to that conversation so take it with salt.