Reflective AI for Career Preparation: Designing and Evaluating LLM-Driven Visualization Systems

The transition from university to professional life is marked by stress, uncertainty, and a reliance on fragmented support tools. While AI-powered platforms for resume analysis and interview simulation have become common, they often suffer from bias, hallucinations, and opacity. This thesis presents Career Compass, a web-based system that integrates large language model (LLM)-driven mock interviews with evidence-linked, interpretable visual feedback to support reflective career preparation. Drawing from formative interviews with students, career advisors, and industry professionals, the system was iteratively designed and evaluated with university students. Results highlight five design principles (productive struggle, adaptive realism, workflow cohesion, motivational visualization, and structured complementarity) that guided system development. This work contributes an empirical account of reflective AI for job preparation and offers design insights for broader applications of LLM systems in high-stakes domains.