Motmans & Partners AI Research Scientist Interview Guide

We've seen a clear pattern in our candidates' experiences at Motmans & Partners: this is not a pure research interview, even for an AI Research Scientist. Multiple candidates reported a LeetCode-style coding bar that showed up early and stayed present throughout the process, alongside research conversations that dug into Ph.D. work, papers, and future research direction. The company seems to care less about polished buzzwords and more about whether you can move comfortably between theory and implementation without losing rigor.

A recurring theme is how much weight they place on defending your technical choices. The research talk was described as a real test of whether candidates could communicate to other researchers, not just present slides. In the technical discussions, interviewers asked about statistical estimators, VAEs, batch normalization, and even ML system design prompts like harmful-content detection or location ranking, which suggests they want people who can reason from first principles and connect that reasoning to product constraints. Our candidates also noted that the interviewers were often highly technical, with strong statistics backgrounds, so shallow ML familiarity tends to stand out quickly.

The non-obvious make-or-break factor here is breadth. Candidates who expected a standard research chat were surprised by the amount of coding and system design, while those who over-indexed on algorithms without enough research depth also struggled. The strongest signal appears to be a candidate who can be credible in both worlds: someone who can write clean code under pressure, explain research decisions clearly, and show they understand how an ML idea would actually ship in a product setting.