Generic practice was the problem
Most interview question lists are broad enough to fit everyone and specific enough to help no one. I built InterviewAI to make practice reflect the candidate's own experience. The platform analyzes a resume, generates relevant questions, runs a timed simulation, and gives feedback across several dimensions. The intent is not to predict a hiring decision. It is to help someone notice where their story becomes unclear.
Grounding questions in the resume
The resume is converted into structured experience, skills, projects, and achievements. Questions can then connect to specific claims instead of inventing a fictional background. The system includes both technical and behavioral prompts and varies follow-ups based on the answer. Clear grounding is important because a personalized experience quickly loses trust if it asks about skills the candidate never listed.
Designing feedback people can use
A single score does not explain how to improve. I organized feedback around relevance, structure, specificity, technical depth, and communication. Suggestions point to moments where an answer needed evidence or a clearer result. Timed sessions add realistic pressure, but users can review the transcript and repeat targeted areas instead of repeating the entire interview.
What I learned
InterviewAI taught me to be careful when AI evaluates people. Feedback should be transparent, limited to the submitted response, and framed as coaching rather than truth. The system should never infer protected traits or promise hiring outcomes. The best result is not a candidate who sounds generated. It is a candidate who can explain their real work with more confidence and precision.