CheetCode

CheetCode was inspired by how different real interviews feel compared to typical coding practice. Most platforms only evaluate the final answer, but interviews focus on how you think, explain your approach, and handle being stuck. We built a real-time interview simulator where users code in a live editor while talking through their reasoning. An AI interviewer watches the code in real time, listens to their explanations, and responds like a human: asking follow-up questions and giving light nudges in the right direction without revealing solutions.

One of our biggest challenges was making the interviewer feel natural, since small prompt changes could drastically affect behavior. We used Keywords AI for prompt versioning, model routing, logging, and evaluation so we could control and debug interviewer behavior instead of treating it like a black box. Next, we want to add more interviewer personalities and deeper feedback as well as hands free conversation to make interview prep feel as close to the real thing as possible.