The soft skills and architecture rounds were solid - fair questions, professional interviewers, real conversations. The two architecture sessions covered a web crawler (~1M users/day) and a survey sending system (~5k users/day) - relatively modest scale, nothing too exotic.
The coding round was a different story. The interviewer was adversarial from the start, repeatedly stopping me mid-implementation to question whether I actually understood what I was writing. When I approached the optimal solution directly - based on patterns from their own prep materials - it was treated as suspicious rather than as preparation paying off. The constant interruptions made it impossible to get any flow, and the whole thing felt less like a skills assessment and more like an interrogation.
The process is well structured overall, but the coding round experience seems to depend heavily on who you get.