InterviewAI

InterviewAI is an AI-powered interview preparation platform: candidates pick a role and a difficulty level, get LLM-generated technical questions, and submit written or voice responses that are scored against a rubric by the same model. The scoring is rubric-driven (per role, per seniority) so the same answer scores differently for a junior vs a senior candidate.

The stack is React 19 on the frontend, Express on the backend, MongoDB for the question bank and user history, and the OpenRouter API (DeepSeek model) for generation and scoring. Question generation is parameterised by role taxonomy so the bank grows predictably and stale questions can be retired without breaking the rest of the tree.

The interesting design choice was making the rubric the primary unit, not the question. Each role has a hand-written rubric with weighted criteria; questions are sampled from a per-role pool. New roles ship without a code change, just a new rubric file.

Open the repo