EduPulse AI

Student Productivity, AI Study Co-Pilot & Learning Orchestration

Comment

EduPulse AI: Academic Intelligence Platform

Track: Student Productivity, AI Study Co-Pilot & Learning Orchestration

Problem

Students face severe cognitive overload due to fragmented study workflows. Learning materials are scattered across multi-hour YouTube lectures, lengthy PDF textbooks, disjointed schedule planning apps, and unintegrated timers. Traditional study approaches result in poor retention, wasted study hours, and inefficient preparation for exams.

edu1

The Solution: EduPulse AI

EduPulse AI is a unified, multi-model academic co-pilot engineered to transform raw educational resources into actionable knowledge. Powered by an ensemble of Google Gemini API (for complex reasoning, synthesis, and roadmap generation) and Hugging Face Inference Models (for specialized NLP, entity extraction, and fast local/cloud summarization), EduPulse AI provides:

  1. YouTube Video Intelligence Engine: Extracts transcripts, generates timestamped structural notes, flashcards, and automated self-assessment quizzes.
  2. Multi-Modal PDF & Document Synthesizer: Parses textbooks and lecture notes into structured summaries, key concept maps, and context-aware Q&A.
  3. Adaptive AI Roadmap & Schedule Generator: Builds customized, goal-driven study roadmaps complete with task breakdowns, difficulty estimations, and milestone checkpoints.
  4. Interactive Focus Timer & Analytics Engine: Integrates an animated Pomodoro/Focus timer with real-time study analytics and audio cues to optimize focus states.

Target Users

  • University and Secondary Students: Seeking structured review materials and efficient study workflows.
  • Self-Directed Learners: Processing technical documentation and video courses.
  • Educators: Extracting quick assessments and active recall decks from lecture material.

⚠️ Main Challenge

The principal technical challenge was enforcing strict structural output integrity from large language models across diverse user inputs. Handling video transcripts that exceed model context windows, managing YouTube videos with disabled captions, and preventing API failure during invalid client requests required strict Pydantic validation schemas, server-side payload truncation, and resilient error propagation to the client interface.

Track Alignment

Education Technology and Generative AI Track.

edu2

Impact

EduPulse AI reduces preparation overhead for study sessions and improves active recall practice. Users can break down long video lectures and notes into verifiable takeaways, practice quizzes, and structured daily study schedules.

Key Results

  • Transcript Extraction & Processing: Automated transcript retrieval and generation of executive summaries, core concepts, and context-aware quizzes from YouTube URLs.
  • Active Recall Deck Generation: Created a structured pipeline that converts raw text inputs into multi-card flashcard decks with dynamic front-and-back rendering.
  • Timeline Planning Algorithm: Developed an endpoint that calculates daily hour allocations and subject roadmaps based on user-defined constraints.
  • Unified Interface: Built a lightweight client interface that toggles workspace tools without page reloads.

🛠️ Tech Stack

  • Backend Framework: Python 3.11, FastAPI
  • Data Validation: Pydantic v2
  • Language Models & APIs: Google Gemini API (google-generativeai), Hugging Face Inference API
  • Integrations: youtube-transcript-api
  • Frontend Architecture: Async Vanilla JavaScript (ES6+), HTML5, Tailwind CSS
  • Server Execution: Uvicorn ASGI

Prerequisites

  • Python 3.11 or higher
  • Python pip package manager
  • A valid Google Gemini API key
  • A valid Hugging Face API key
Share this project:

Updates

Submission history