AI Interview Trainer Agent — Personalized Mock Interview Coach with IBM Granite & RAG
Link to open source: https://github.com/Pranavv28/Confluence-AI
An agentic AI system built on IBM Granite (via Watsonx.ai) that conducts personalized mock interviews. Given a candidate's job role, experience level, and domain, it generates 3 tailored technical questions and 2 behavioral (STAR-format) questions, then evaluates each answer in real time — scoring it out of 10, flagging strengths and gaps, and providing an improved model answer. Questions are grounded using RAG over a curated interview Q&A corpus (Chroma vector store), and the agent maintains conversation memory across the full 5-question session, adapting follow-up difficulty based on prior responses. The goal is to close the gap left by generic interview-prep tools, which give one-size-fits-all questions with no real-time feedback — leaving candidates unprepared and under-confident going into real interviews.
This build was uploaded as a hackathon project
