InterviewExceler.AI: An AI-Powered Platform For Mock Interviews, CV Analysis, and Real Time Behavioural Feedback

Abstract

Preparing for real-world interviews often requires multiple disconnected tools for practice, CV editing, and performance feedback. To bridge this gap, we present InterviewExceler.AI, a full-stack platform that provides an integrated and realistic mock interview experience. The system employs transformer-based models (BERT, SBERT) for role-specific question generation and client-side facial landmark analysis using CNNs with TensorFlow to capture non-verbal behaviour. These inputs are combined with heuristic scoring and Gradient Boosted Trees (XGBoost) to generate structured and actionable feedback.

The platform integrates an interactive Next.js frontend, serverless convex functions, and modular ML pipelines, ensuring scalability and adaptability across diverse interview formats. Designed with reproducibility and extensibility in mind, it provides architecture details, implementation guidelines, evaluation protocols, and open-source artifacts to enable further research and adoption.

Evaluation highlights effectiveness in three areas: generating relevant interview questions, analysing multimodal responses, and offering personalized coaching recommendations that improve with repeated use. Beyond technical contributions, InterviewExceler.AI seeks to reduce bias, enhance accessibility to professional interview training, and lay the foundation for next-generation AI-driven career support systems.