Setting Page
Assessment Page
Study Planner Page
AI Tutor Page
Dashboard Page
Course Page

Inspiration

What it does

OmniTutor AI: Personalized Tutoring & Adaptive Multimodal Learning Platform

Grounded Socratic scaffolding, verifiable provenance citations, anti-answer-leakage assessments, and mathematical forgetting-curve calibration.

💡 Inspiration

Generative AI has ignited a revolution in self-guided education, but today’s mainstream AI study tools suffer from two critical pedagogical flaws:

Hallucination & Lack of Grounded Provenance: Generalist LLMs frequently hallucinate facts, equations, and textbook citations. When a student is preparing for high-stakes university exams, a plausible-sounding hallucination can be disastrous.
Answer Spoon-Feeding: When asked for homework help, off-the-shelf chatbots immediately dump the final answer. This deprives students of the cognitive struggle required to build long-term conceptual intuition.
Assessment Integrity & Cheating: Most web quiz platforms inadvertently send answer keys to the client browser, making exam cheating as simple as opening Chrome DevTools.

We set out to build OmniTutor AI—a tutor that doesn't just give answers, but guides students toward them. We wanted a system that grounds every explanation in actual university syllabus materials, verifies citations interactively, secures assessments against answer leakage, and models memory retention using cognitive science.

🎯 What It Does

OmniTutor AI transforms raw course materials into an interactive, trustworthy learning ecosystem:

Multimodal Course Ingestion: Students and educators upload lecture slides, syllabi, PDFs, DOCX, and lecture audio/video. The platform validates file integrity, extracts text, and indexes content into a high-dimensional vector database.
Socratic Tutoring Workspace: Instead of answering directly, the AI adopts pedagogical personas (❓ Probe, 💡 Hint, 🧩 Synthesize, 📖 Explain) to prompt student self-discovery.
Interactive Provenance Citations ([cite:N]): Every claim made by the tutor includes a clickable badge that instantly opens a provenance popover displaying the exact excerpt, source document, and confidence score.
Anti-Leakage Adaptive Assessments: Quizzes dynamically calibrate difficulty based on student performance. Crucially, answer keys, explanations, and grading rubrics are physically stripped from API payloads until submission.
Ebbinghaus Forgetting-Curve Scheduling: Based on spaced-repetition cognitive heuristics, the system predicts when memory decay will occur and automatically schedules targeted revision sessions.
Knowledge Gap Detection: The platform aggregates assessment mistakes across cohorts, isolating recurring misconceptions and recommending specific source paragraphs for review.