Exam Buddy is a locally operating AI study tool designed to transform personal notes into interactive quizzes. The application aims to improve learning by shifting from passive re-reading to active recall, a more effective study method. It generates quiz questions directly from the user's provided study material, ensuring relevance to their specific course content.When users answer incorrectly, Exam Buddy offers personalized explanations tailored to their preferred themes, such as sports, cooking, or movies. The tool provides multiple quiz modes, including multiple-choice, AI-graded written answers, and a mixed format. It incorporates a follow-up question-and-answer feature for deeper understanding and a "retry-missed" mode to focus on challenging topics.Exam Buddy also tracks score history and generates a summary of areas needing review after each quiz. A significant aspect of the application is its fully offline operation, utilizing Ollama and the Gemma 3 4B model. This ensures that users' notes and data remain on their local machine, enhancing privacy and security.The development involved overcoming challenges with small model reliability, such as inconsistent JSON output and prompt engineering for accurate grading. The developer also addressed UI-related issues like answer shuffling and streaming data race conditions. The project highlights the practicality and power of smaller, locally run AI models for personalized educational tools. Exam Buddy's design prioritizes user privacy by avoiding reliance on external cloud APIs for AI processing. The project demonstrates that sophisticated AI tasks can be effectively handled by modest, local models with proper implementation and guardrails.