Cogniflow - Offline LMS

Education That Flows Beyond Connectivity.

Comment

Inspiration

In many regions, especially rural and underserved areas, access to reliable internet remains a major barrier to quality education. Existing Learning Management Systems (LMS) rely heavily on connectivity — leaving millions of students disconnected from modern learning. We were inspired by a single idea: “Education should not stop when the internet does.” That’s how Cogniflow was born — an AI-powered offline LMS that ensures every learner, regardless of location, can continue their educational journey.

What makes Cogniflow unique:

  • Offline AI Inference: Runs ML models locally using Ollama — no server required.
  • Multi-Modal Interaction: Supports text, speech, and gesture-based learning for accessibility.
  • Edge-Based Personalization: Generates adaptive learning paths locally on each device.
  • Teacher Dashboards: Empower educators to track and manage learning — even without internet.
  • No Cloud Fees : no hosting required for usage
  • Affordable Hardware : Low end GPUs and even CPUs can run the content Cogniflow is not just another LMS — it’s a self-sustaining ecosystem of learning, designed for connectivity deserts.

Discover the Vision Behind Cogniflow

What it does

Cogniflow is an AI-powered, offline-first Learning Management System (LMS) that allows both teachers and students to experience modern education without requiring continuous internet access.

  • Personalized learning modules and quizzes.
  • Offline teacher dashboards for tracking progress.
  • Accessibility through Voice, Text and Gesture Based Interaction
  • Adaptive difficulty levels powered by on-device AI.
  • Leaderboard Management

How we built it

We developed Cogniflow using a modular, offline-first architecture, designed for scalability and low resource consumption.

  1. Quizzes and Flashcards: The Pdf content is uploaded by teachers or students. Dynamic Content is generated with the help of Gemma Model by extracting text from the pdf
  2. Offline Video Generation on Chapters: The video is generated from the script given by Gemma model for extracted text from the chapter.
  3. AI assistant: RAG Based AI Assistant that guild student and solves doubts on chapter wise content. Uses FAISS Embeddings to store the student interactions in Long term.
  4. Offline Functionality: uses Service Worker for offline caching of static and dynamic assets.
  5. Mobile and cross Platform Compatibility: uses Progressive web apps and IP Sharing for Mobile and cross Platform Compatibility.

Output Efficiency (E)=f(Inputs,AI Intelligence,Engagement) \( \text{AI Intelligence} = \text{Google Gemma3n Core} \), \( \text{Engagement} = { \text{Gamified Modules}, \text{Rewards}, \text{Tracking} } \) Thus, $$ E = \alpha (\text{Personalization}) + \beta (\text{Adaptivity}) + \gamma (\text{Offline Sync Accuracy}) $$ with weighting coefficients \( \alpha, \beta, \gamma \) derived empirically.

Challenges we ran into

  • High GPU/VRAM demand : model quantization and Optimization for low end GPUs
  • limited device availability : LAN based and IP based sharing for multi student access.
  • Need for Teacher Control : Teacher Dashboards
  • Maintenance Of Installation : Easy to read documentation for easy Installation.
  • Language & Disability barriers : Use of sign language recognition and speech-to-text enginer.

Accomplishments that we're proud of

  • Built a teacher dashboard that works offline and syncs intelligently when online.
  • Created a modular AI pipeline capable of adaptive content recommendation offline.
  • Integrated speech and gesture recognition into the LMS interface.
  • Achieved efficient local data caching and synchronization flow.
  • Designed Cogniflow to run on low-resource devices without compromising performance.

What we learned

  • How to design offline-first AI systems that scale to real-world conditions.
  • The power of TypeScript + Python hybrid architectures for full-stack AI applications.
  • How to build teacher-centric dashboards with meaningful analytics even offline.
  • Techniques for efficient edge inference and model compression.
  • The importance of UX design for accessibility and inclusivity in education tech.

What's next for Cogniflow - Offline LMS

  • Support for multi-language and regional content delivery.
  • Add peer-to-peer local data sharing between nearby devices for group learning.
  • Introduce gamified learning journeys to improve engagement.
  • Integrate federated learning to let edge devices improve models collaboratively.
  • Expand teacher dashboards with predictive analytics and AI tutoring assistance.

Built With

Share this project:

Updates