Inspiration
The demands for personalized learning tools compelled us to create a tool that would be able to complement traditional math education for small learners. Furthermore, we want to bridge the observed gaps in the understanding of basic arithmetic concept, such as addition, subtraction, multiplication, and division, through personalized exercises tailored to each learner's pace. We also dream of making this tool available to students through local libraries, schools, and tutors to ensure that they are well-grounded in mathematics.
What it does
Our project consists of specialized worksheets designed to improve basic arithmetic skills. These worksheets are powered by an AI model that tailors the exercises to the student's level. By using Retrieval-Augmented Generation (RAG), the system generates targeted exercises to ensure each student gets the practice they need. These worksheets are intended for distribution through schools, libraries, and tutoring sessions to supplement classroom learning.
How we built it
We use RAG, training an AI model on a database of written arithmetic exercises to create personalized learning materials for students. On the back end, technical work involved combining the RAG model with JavaScript to generate and display the exercises in an intuitive format. The structuring of the exercises was targeted so they can be educative and engaging.
Challenges we ran into
Our biggest challenge was the training of the RAG model and how it would seamlessly work with JavaScript in order to deliver these exercises in a dynamic way. Second, model finetuning for different levels of difficulty and optimizations on how it retrieves relevant exercises also created challenges. The coordination of those technical elements has become a key focus of our efforts.
Accomplishments that we're proud of
We’re proud of how engaging and accessible the exercises have become, thanks to the flexibility provided by the AI model. We have also learned a lot about working with vector embeddings and applying them to optimize the performance of our RAG model, ensuring that it delivers exercises suited to each student’s current level.
What we learned
This project gave us deep insights into how vector embeddings work and how they can be applied to educational content. We also learned about the technical challenges of integrating an AI model with a front-end interface, particularly in using RAG and JavaScript together to ensure smooth and accurate generation of exercises.
What's next for La Trinidad
Our next steps include gathering more data to improve the AI model and fine-tuning the RAG system for even better performance. We also plan to expand our content library and refine the model’s ability to adapt to various learning speeds. Scaling distribution through more libraries, schools, and tutoring services remains a top priority as we continue to build partnerships and enhance the user experience.
Log in or sign up for Devpost to join the conversation.