Friday : An AI study companion

STARTING APP ON SCREEN WINDOW
COURSE SELECTOR WINDOW
LANGUAGE AND STUDY MANNER SELECTOR FOR MAKING STUDY MORE EASY
PACING AND CONFIDENCE LEVEL WINDOW
MAIN APP DASBOARD
OFFICIAL NEWS AND UPDATE FEATURES OF THE EXAM SELECTED BY USER
STUDENT ERROR BOOK WINDOW FOR STORING ALL ERROR OF THE EXAM DONE BY USER
FRIDAY AI STUDY COMPANIAN CHATBOT WINDOW
REGULAR MOTIVATION LINE TO MAKE MOTIVATION MORE HIGH
EXAM SIMULATOR WINDOW FOR CUSTOM EXAM PRACTICE
AI STUDY CAMPANIAN ANALYSIS OF OVERALL SUBJECT BASED ON TEXT
AI STUDY PLANNER BASED ON THE WINDOW SELECTED BY USER
VIRTUAL SIMULATOR TO MAKE LEARNING EASY
FLASH CARD , MEMORY CARD , FORMULA SHEET FOR QUICK REVISION
VIDEO AND LECTURE NOTES SUMRAZIER

InThe inspiration for Friday came from the "fragmentation gap" in

modern education. Students are overwhelmed by a chaotic mix of 2-hour lecture videos, 500-page PDFs, and scattered slide decks. We realized that while general AI chatbots exist, they lack the specific context of a course and often hallucinate. We wanted to create a "Digital Brain" that doesn't just give answers, but acts as a personalized tutor that knows exactly what is in your syllabus and exactly what you have yet to master. We built Friday using a sophisticated AI pipeline:

The Knowledge Base: We implemented a RAG (Retrieval-Augmented Generation) pipeline capable of processing multimodal data (OCR for PDFs/Slides and Transcription for Videos).
The Intelligence: We utilized a Large Language Model (LLM) constrained by a "grounding" layer to ensure the AI declines queries that fall outside the provided course material.
The Learner Model: We integrated a tracking system (inspired by Bayesian Knowledge Tracing) to estimate topic mastery based on quiz performance.
The Frontend: A clean, intuitive dashboard built to allow students to upload materials and interact with their "Friday" companion seamlessly. Multimodal Alignment: One of our biggest hurdles was linking a text-based answer back to a specific second in a video or a specific figure in a textbook.
Hallucination Control: Ensuring the AI stayed strictly within the "source-grounded" boundary required rigorous prompt engineering and iterative testing.
Personalization Logic: Building a learner model that accurately reflects "mastery" rather than just "correct
Learner Modeling: It tracks mastery of individual concepts in real-time, adjusting the difficulty and focus of the tutoring session as the student grows. Throughout this project, we dove deep into the world of Multimodal RAG and Adaptive Learning. We learned how to quantify AI "faith comes with a precise citation (timestamp, page number, or slide) to ensure trustworthiness.
Adaptive Assessments: Friday generates personalized quizzes and mock exams based on the student's current weak points.
Learner Modeling: It tracks mastery of individual concepts in real-time, adjusting the difficulty and focus of the tutoring session as the student grows. Friday is just getting started. Our roadmap includes:
Visual Course Maps: Generating an interactive "Knowledge Graph" so students can see a visual map of their learning journey.
Proactive Scheduling: Integrating "forgetting curves" to suggest exactly when a student should review a topic before they forget it.
Audio-First Tutoring: Moving beyond chat to allow full voice-based tutoring sessions for a more natural experience.
Multilingual Support: Enabling students to upload English lectures but receive explanations in their native language (e.g., Hindi).spiration