Multimodal AI Hackathon

• 

  MentorAI Source-Grounded Q&A with Inline Citations and Course Evidence Drawers

• 

  Adaptive Assessment Engine & Real-Time Skill Mastery Diagnosis

• 

  Measurable Educational Progress via Hake's Normalized Learning Gain (g = 0.67)

• 

  MentorAI Closed-Loop System Architecture & Multi-Format Parsing Pipeline

About the Project

MentorAI is an evidence-first, closed-loop AI study companion engineered for Track D: Personalized Tutoring and Adaptive Learning.

While traditional AI study tools act as passive, single-turn chatbots that output answers without verifying comprehension, MentorAI closes the learning loop across five key stages:

• Multi-Format Ingestion: Ingests textbook chapters (.pdf), professor lecture slides (.pptx), and video transcripts (.srt/.vtt) into a unified knowledge base.
• Source-Grounded Q&A: Synthesizes structured explanations backed by direct inline source citations and verifiable course evidence drawers.
• Explainable Learner State: Tracks real-time student mastery across granular sub-skills in an intuitive dashboard.
• Targeted Micro-Remediation: Diagnoses specific weak skills upon quiz errors and triggers instant remedial micro-lessons.
• Measurable Progress: Quantifies tangible educational improvement using Hake's Normalized Learning Gain Formula (g = 0.67).

Built With

• adaptive-learning
• ai-tutor
• data-parsing
• edtech
• educational-analytics
• generative-ai
• hake's-normalized-gain
• interactive
• katex
• latex
• machine-learning
• multimodal-ai
• natural-language-processing
• numpy
• offline-first
• openai
• openrouter
• pandas
• personalized-learning
• pypdf
• python
• python-pptx
• rag
• streamlit
• vector-retrieval