Simplified Homework AI

Brainstorming Ideas
Outlining Project
Overwhelmed by Errors
Coding
Code Works!
Finished Everything!

Inspiration

Sometimes we struggle with our homework, and we understand that other people struggle with homework like we do. So we dedicate our project to effectively guide you through your homework

What it does

Our project is an AI-powered homework tutor that helps students understand and solve problems step by step- just like a real teacher would. Here’s how it works: Snap a Photo: The student takes a picture of their homework using the webcam. Text Extraction: The app uses EasyOCR to read and extract the question from the image. Homework Detection: The AI determines whether the text is an actual homework problem or just general text. Smart Tutoring: Once confirmed, the AI identifies the subject and topic (like algebra or physics), explains key concepts, asks checkpoint questions, gives feedback, and adjusts its teaching style based on the student’s responses. In short, it’s a personal AI tutor that can see, read, think, and teach - all within one interactive Python application.

How we built it

We coded this project entirely in Python, leveraging several key libraries to bring each component to life. To integrate large language models (LLMs) directly into our application, we used Ollama, which allowed us to locally run models such as Gemma 3 (4B) for fast and private AI responses. We combined LangChain and Ollama to handle structured prompt templates and response chaining for different tutoring tasks (like classifying homework, explaining concepts, and providing feedback). For image input, we used OpenCV to capture photos from the webcam and EasyOCR with Pillow (PIL) to extract the text from homework images.

Challenges we ran into

Since this was our first hackathon, we didn’t really know what to expect and weren’t fully prepared for everything that came our way. One major challenge was our limited experience with web development, so we needed to come up with a python alternative: Streamlit. We also faced a steep learning curve as we had to pick up new tools and frameworks on the fly, including Ollama, LangChain, and EasyOCR. Despite the difficulties, we kept learning throughout the process and successfully built a working prototype that exceeded our initial expectations.

Accomplishments that we're proud of

We were able to overcome our lack of design skill by using a python alternative
That we completed and finished our first hackathon
That we made a functioning project
That we spent a lot of time and effort onto this
And how we learned a lot from this

What we learned

That we should be more prepared for our next hackathon
How to implement OpenCV for image capturing, and Ollama for integrating large language models into our project