Tiger Analytics AI Engineer Interview Guide

We've seen Tiger Analytics lean hard into whether candidates can connect AI concepts to real delivery, not just recite definitions. In our candidate experience, questions around temperature settings, prompt caching, and evaluating a RAG system came up alongside AWS and prompt engineering, which tells us the bar is less about academic polish and more about whether you understand how these systems behave in production. A recurring theme is that the interviewers want practical judgment: when a model should be more or less creative, how retrieval quality is measured, and what tradeoffs show up once an LLM workflow is actually deployed.

The other signal that stands out is how much weight they place on your own work. Multiple candidates reported that the conversation became much deeper once the team started probing project details, and that’s where the interview felt most demanding. We’d read that as a strong preference for ownership and technical specificity — if you built it, you should be able to explain the architecture, the failure modes, and why you made each choice. There’s also some unevenness in the process, with a few questions drifting into transformer math or less relevant coding, so the safest candidates are the ones who can stay grounded in fundamentals while still steering back to applied AI reasoning.