MotionMind

Inspiration

In traditional educational settings, many students struggle to keep up, not due to a lack of intelligence, but because their unique learning needs and preferences are not met. This is particularly true for those with ADHD, dyslexia, or autism, who may thrive with alternative learning methods. MotionMind was inspired by the need for a more inclusive, accessible, and tailored educational experience for all learners, regardless of background or ability.

What it does

MotionMind provides a personalized and inclusive learning experience using AI-powered visual aids. By transforming complex concepts into engaging, animated visuals, it helps students—especially those who benefit from visual learning—understand and retain information more effectively. MotionMind is designed to support diverse learning preferences, making education more accessible and engaging.

How I built it

I built MotionMind with a robust tech stack, including Next.js for the frontend, Flask for the backend, and OpenAI's API for generating content. The AI models were fine-tuned using a JSONL dataset to create more accurate and tailored responses. Additionally, I used Manim and FFmpeg to generate animations based on user input, making the content visually engaging. Using Cerebras Inference, I enhanced the speed of our large language model, achieving performance benchmarks approximately 16 times faster than without this optimization.

Challenges I ran into

• AWS EC2: One of the main challenges I encountered was setting up a smooth CI/CD pipeline on AWS EC2.
• Model Fine-Tuning: Adjusting the AI model to handle specific queries for educational content was challenging, especially given token limitations and processing time.
• Rendering Animations: Integrating Manim to generate real-time animations based on user queries posed technical challenges, particularly with ensuring smooth transitions and accurate representations.

Accomplishments that I'm proud of

• Successfully fine-tuning a model to generate educational content tailored to diverse learning styles.
• Creating a seamless text-to-video pipeline that dynamically generates animated content based on user input.
• Building an inclusive learning tool that makes complex subjects more accessible to visual learners.

What I learned

• The importance of preprocessing data effectively for fine-tuning AI models to ensure accurate outputs.
• Techniques for managing and optimizing rendering processes with Manim to create smooth, engaging visual content.
• Working with serverless APIs and improving my skills backend in general.

What's next for MotionMind

• Enhanced Personalization: I plan to further fine-tune our models to adapt to individual learning styles based on feedback and user interaction data.
• Broader Content Library: Expanding the range of subjects and topics covered, making MotionMind a comprehensive educational tool for various disciplines.
• Mobile and Web App Development: Bringing MotionMind to mobile and tablet devices to make it even more accessible to students on the go.

Built With

• finetunedllm
• framermotion
• manim
• nextjs
• openai
• serverless-api
• tailwind
• vercel