Microsoft MAI Copilot onsite, rejected after the follow ups escalated

Microsoft MAI Copilot onsite, rejected after the follow ups escalated

Just wrapped up my MAI Copilot loop at Microsoft, four rounds, got the rejection email yesterday.

HM round was mostly a project deep dive plus the standard Microsoft leadership stuff, drive for results, bias for action, nothing unusual.

Coding 1 was a decent LC medium with a couple follow ups that kept stacking. Coding 2 is the one worth writing up. You get an array plus a list of query indices, and for each query you're counting how many times the current max of that suffix repeats, from that index to the end.

Brute forcing it means walking each suffix per query, which falls apart once both get large. The trick is one pass from the back, tracking a running max and how many times you've seen it, resetting whenever something bigger shows up. I got the constant time version working but the interviewer kept pushing on negative values and maxes repeating right next to each other, and I fumbled one of those follow ups.

SD was the chatbot memory system, RAG based, handling conflicting history. Felt okay walking in but I'm guessing BQ or that follow up depth is what got me. Back to grinding I guess.