Rejected after Microsoft Applied Scientist IC2 final loop – would appreciate insights on what might’ve gone wrong

Hi everyone, I recently interviewed for the Applied Scientist IC2 role at Microsoft (Teams group) and received a rejection after the final loop. I’m trying to understand where things might’ve fallen s

Hi everyone, I recently interviewed for the Applied Scientist IC2 role at Microsoft (Teams group) and received a rejection after the final loop. I’m trying to understand where things might’ve fallen short and how I can improve for future opportunities. Here’s a quick summary of my interview experience: * Round 1 (

System Design Software Engineer): This was my weakest round. It was an unexpected design interview focused on reducing prompt token usage. I got a bit too detailed early on instead of starting with a high-level plan. The interviewer’s feedback was to “break problems down, start broader, and avoid jumping too deep too fast.” * Round 2 (ML/Math – Principal Applied Scientist): This one went quite well. We discussed CNN architectures and deriving the PCA equation (eigenvalues, eigenvectors). I explained my reasoning clearly, and the interviewer seemed satisfied, even mentioning that they liked my projects. * Round 3 (LLM & Experimentation – Senior Software Engineer): Focused on RAGs, GenAI, A/B testing, and MCP. I was able to explain all concepts confidently, and the interviewer seemed happy with the depth of my responses. * Round 4 (LLM & ML Evaluation – Senior Applied Scientist): This was my best round. We talked about LLM evaluation metrics (BLEU, ROUGE, human evals, LLM-as-a-judge, etc.) and ML evaluation in depth. The conversation flowed smoothly and felt collaborative. Overall, I genuinely felt good about 3 out of 4 rounds. The recruiter later told me “Unfortunately, Microsoft will not be moving forward with an offer. I haven’t been provided any feedback associated with your conversations; however, I’m happy to try and answer what I can outside of that” I’m curious , for those who’ve been through similar loops at Microsoft or similar companies — * Could one average system design round outweigh the rest? * Do such rejections often come down to “fit” or headcount rather than performance? * How can I improve my design-thinking and high-level reasoning for such roles? Would really appreciate any insights or advice from folks who’ve gone through this process. Thanks!