ThinkFlow

Inspiration

The idea for ThinkFlow was born from the need to make learning more structured, accessible, and intuitive. Many learners struggle with information overload, making it difficult to synthesize key concepts effectively. Inspired by cognitive learning theories and visual learning techniques, ThinkFlow aims to bridge this gap by leveraging AI-powered mind maps to simplify complex topics.

What It Does

User personalisation
Summary
Mind Map Generator
Accessibility
1. Text-to-Speech
2. Reading Assistance with special font and size for textual data

Beta Version

Quiz Generator
More personas for personalisation

What We Learnt

Throughout this journey, we gained insights into:

Digital literacy challenges and how AI can support structured learning.
Accessibility barriers in digital education and the importance of assistive technologies like text-to-speech and reading assistance.
Personalized learning models that adapt to different cognitive styles and user personas, enhancing engagement and retention.

How We Built It

ThinkFlow was developed using a combination of:

AI-powered Personalisation using Deepseek and Mistral
Mind Map Generation using Node.js and D3.js
Speech and accessibility tools for inclusive learning experiences using OpenDyslexic font and Google TTS API
MongoDB to store user and mind map data.

Challenges We Ran Into

Information Structuring: Designing an AI model that accurately extracts and organizes key concepts without losing context.
Accessibility Features: Integrating assistive technologies while ensuring a smooth user experience.
Integration: Between Streamlit and React Frontend App

What's Next For ThinkFlow

Our goal is to offer a wide range of learning materials, including videos and articles from trusted sources across the web, enabling learners to explore content in diverse formats.

Key features include:

Interactive Learning Environment
Personalized Learning Pathways
Accessibility for All
Collaborative and Community Learning

Despite these challenges, ThinkFlow is a step toward democratizing digital learning, making structured knowledge accessible to everyone. 🚀

Built With

css
d3.js
deepseek
express.js
html
huggingface
llm
mistral
mongodb
nextjs
node.js
ollama
python
react
react.js
streamlit

Submitted to

Civic Tech Hackathon 2025
- Winner Best Impact

Created by

I worked on developing personalized model responses using DeepSeek-R1 on Ollama, enabling user-tailored interactions. I implemented LLM-based mind map generation by creating JSON structures based on user input. To enhance personalization, I designed three distinct prompts for different user personas to customize the learning experience.

On the backend, I set up a MongoDB database for data storage and integrated it with Streamlit outputs using PyMongo. Additionally, I developed the Streamlit UI, allowing users to input their details and topics seamlessly, ensuring a smooth and interactive experience. Also worked on integration between Streamlit and React.js app

Ritika Kumar
Developed a question-answer generation system for flashcards using Ollama (Mistral model) to ensure dynamic and relevant content.
Implemented a text-to-speech feature, enhancing accessibility by enabling spoken narration of flashcards.
Assisted in backend integration, ensuring seamless data retrieval and API connectivity for an improved user experience.
Supported frontend enhancements, improving UI/UX elements such as layout adjustments, styling refinements, and accessibility features.

nidhi patel
I worked on training an LLM using Ollama and Mistral, assisting in model fine-tuning and deployment. Additionally, I contributed to developing accessibility features in the frontend, specifically designed for users with reading disabilities and text speech. My role also involved frontend development and supporting the final project presentation.

Suchita Sharma
I contributed to the development of accessibility features on the frontend, focusing specifically on the text-to-speech functionality and providing reading assistance through customizable fonts and text sizes. Additionally, I supported both the frontend and database aspects of the project and worked collaboratively with a teammate to create the final project presentation.

Bhoomika Gupta
Currently a CS graduate student at Northeastern University
I've focused on React.js and D3.js to develop a dynamic mind map graph for the frontend. I integrated accessibility features such as the OpenDyslexic Font. Additionally, I designed and implemented Flashcards and the modal component. On the backend, I developed a Node.js API with MongoDB to ensure seamless data integration with the frontend. I also worked on creating and integrating the Streamlit app with the React application.

Shivani Sharma