Inspiration

We aimed to leverage to power of Intel's AI PC's lives after considering various in education, we decided to create an AI Tutor chatbot to assist UW students with academic resources, exam review, and one-on-one tutoring available anytime.

What it does

Ideally, you would be able to give the application exam questions and your answers, and the application would be able to generate a review exam based on your answers and your weak points. The application would then help you by giving you tools and guidance without giving straight answers so that users can learn the topics, not just solve them.

How we built it

We integrated the Ollama API into Python, and supplied queries in domains of Biology, Chemistry, Philosophy, Politics, & Economics. Our demo currently uses biology queries, enabling it to fully answer biology related questions.

Challenges we ran into

We had trouble selecting the appropriate pre-existing AI model to use, and implementing said model into Python. Additionally, finding ways to fine-tune the model for providing accurate and helpful responses required considerable effort, as we want the AI to encourage thinking, and not give clear answers.

Accomplishments we are proud of

We successfully developed a prototype where Ollama responds through our program instead of command line. This achievement allows us to customize the AI model for particular use cases, e.g. study tutor for biology.

What we learned

We learned that AI is difficult but rewarding to work with. We gained experience in model selection and integration, along with fine-tuning the model, which will be beneficial for our future projects.

What's next for Learning2Learning

Gather more data specific for biology and sort it by topics, subtopics, questions, and answers. Later we want to incorporate multiple data sets so that our AI tutor is not only limited to a single subject. We aim to create a versatile AI Tutor that can assist students in various academic studies.

Built With

Share this project:

Updates