Winston

Inspiration

We wanted to create Winston to give students a way to prepare for upcoming quizzes or tests in classes where practice material may not be provided by instructors.

What it does

Winston uses AI to create practice questions that test a user's knowledge on the contents of any given PDF. Winston can generate multiple choice, open-ended, and true/false questions.

How we built it

We used React and Tailwind.css to create our UI. We used Flask for our backend development. To generate the practice questions, we used OpenAI's API. To grade the open-ended questions, we use sklearn and cosine similarity.

Challenges we ran into

Some challenges we ran into include receiving the responses from OpenAI's API in a consistent format, grading the user's responses to open-ended questions, and using conditional formatting to show the results of a user's submission.

Accomplishments that we're proud of

The accomplishments we are most proud of are using AI to help students in a meaningful way and participating in our first hackathon.

What we learned

Two things we learned were backend development via Flask and using specific prompts with AI to generate reliable responses. Also, since this was our first hackathon, we learned how to balance between building our project and meeting others who are interested in building.

What's next for Winston

Some ways that we would like to expand Winston to a wider audience and increase its usability is to give users more options in regards to what file types they can upload (audio, video, etc), to allow users to save difficult questions if needed for review, and to direct students to dependable resources if any extra explanation is necessary for incorrect answers.

Built With

• flask
• openai
• pdfplumber
• python
• react
• sklearn
• tailwind.css
• typescript