Project Name: Guru Cortex
Project Overview Project Cortex is a personalized, proactive AI study buddy designed to fundamentally change how students interact with educational AI. Unlike most AI tutors that simply wait to answer questions, Cortex utilizes a proactive approach—it actively pays attention to the student, notices when they are confused, and offers assistance before they even have to ask.
Core Problem & Solution Current educational AI relies on reactive prompts. Project Cortex solves this by acting as a highly observant tutor that provides affordable, accessible, and localized learning. It achieves this through a Privacy-First Hybrid Architecture (Edge + Cloud), ensuring that sensitive vision processing happens locally on the user's device while complex reasoning is handled securely.
Key Features
Smart Vision System & Emotional Check-ins: Using a webcam, the system monitors for gaze, fatigue, and frustration. It detects when a student is confused or tired and proactively suggests new approaches or study breaks.
Local Knowledge Base (DeepSearch): Instead of relying on generic internet data, Cortex indexes the student's personal PDFs and notes. This allows it to provide highly contextual answers based on the user's actual study materials.
Active Learning Mode: Features a "Teach Mode" that prompts the student to explain concepts back to the AI to verify and enhance comprehension.
Multilingual Support: Offers conversational AI capabilities in Hindi and other regional languages.
Drawing & Diagrams: Automatically generates visual flowcharts to help simplify complex text.
Focus Features: Detects distraction and actively alerts the student to stay focused on their screen.
Technology Stack & Architecture The system is built on a Hybrid Edge-Cloud model to balance performance with strict privacy:
AI Models: Powered by Quantized Google Gemma (2B/9B) for language processing and Whisper Tiny for efficient voice input and output.
Vision & Emotion Processing: Utilizes MediaPipe and OpenCV locally on the device, ensuring 100% video privacy without sending visual data to the cloud.
Storage & Memory: Employs ChromaDB for a Local GraphRAG vector store (to search notes) and SQLite to track student progress and spaced repetition.
Interface & Orchestration: Features a Next.js dynamic dashboard with Framer Motion UI, orchestrated by a Django/FastAPI server hosted on Google Cloud Run. Mermaid.js is used for generating diagrams.
Future Development Roadmap
Multimodal Learning: Integrating Multimodal Gemma to analyze diagrams from textbook photos and digitize handwritten notes via OCR.
Cross-Platform App: Transitioning to Flutter for seamless desktop-to-mobile syncing and running Gemma models locally on mobile devices.
Privacy-Safe Study: Enabling secure collaborative learning between students using peer-to-peer (P2P) encryption and the ability to share Knowledge Graphs without sharing raw documents.
Built with