Claude Builder Club Northumbria
LectureMate
Participants
Arthur Wheildon
Theme
AI for Personalised Education
Live Demo
Coming Soon
Repository
View CodePresentation
View Slides
LectureMate is an AI-powered educational tool that transforms complex lectures into personalized summaries tailored to each learner's expertise level. Transform complex lectures into personalized summaries tailored to your expertise.
Solution
LectureMate features five distinct adaptive knowledge levels from Complete Beginner to Expert, each using different language complexity, technical terminology, and depth of explanation. The Multi-Source Memory Bank combines multiple inputs into one comprehensive lecture, supporting text, PDF, and audio uploads with session-based memory. Real-time streaming summaries generate live with a character-by-character typewriter effect and instant level switching. Intelligent document processing handles PDFs, Word documents, PowerPoint presentations, and plain text with automatic extraction and formatting.
Challenges
Traditional lecture materials are one-size-fits-all, causing beginners to feel overwhelmed with jargon, experts to waste time on basic explanations, and students to struggle finding the right learning depth. LectureMate solves this with audio transcription using Whisper AI supporting 10 languages, an interactive Q&A system with context-aware responses tailored to knowledge level, and an adaptive quiz system with dynamic difficulty adjustment that finds each student's true knowledge ceiling through natural language evaluation.
Results
Built with Vanilla JavaScript and Server-Sent Events for real-time streaming on the frontend, FastAPI (Python) on the backend, Anthropic Claude Sonnet 4.5 for AI reasoning, and OpenAI Whisper for audio transcription. Key achievements include real-time streaming with character-level precision, session-based memory management, parallel summary generation across 5 levels simultaneously, support for 10 languages and 5 file formats, handling PDFs up to 50MB and audio up to 200MB, and a mobile-responsive single-page application with graceful error handling.
