Inspiration
We were inspired by our AP Calculus BC teacher, who explained to us that she works tirelessly, staying 2 and a half hours late after school every day to grade work and create assignments. This time is unpaid, and leaves her with less time with her family. She explained to us that her other option is to spend $300 on Teachers Pay Teachers for a lesson plan and structured assignments, a price she does not feel is worth the money. Teachers are stuck choosing between spending money or time on their students. Seeing this hole that needed to be filled, we created GradePilot. With GradePilot, assignments, tests, and quizzes can all be made on one platform, then easily scanned and graded, with detailed explanations on errors and common misconceptions throughout the workflow. GradePilot will save teachers an amazing amount of time, and allows teachers to stop compromising between time and money.
What it does
GradePilot is a website where teachers can
Create assignments and rubrics by writing a simple prompt into our AI tool. This will allow teachers to spend less time creating assignments, and more time collaborating and assisting students in meeting their goals.
GradePilot allows teachers to scan large amounts of assignments within the app itself and our AI will grade each individual assignment, along with finding common misconceptions and patterns of errors throughout the submitted work.
How we built it
Frontend: Built the interactive app using Python and Streamlit so teachers get a clean, simple dashboard to view grades and analytics.
AI Features: Connected Hugging Face AI models to help create rubrics, grade student work, and write helpful feedback automatically by analyzing and generating images of assignments and student progress
Backend: Used VS Code to generate the tool, edited it by testing on our browser, and stored and updated all our project code on GitHub
Testing: Ran automated tests several times and manually used the tool to make sure the pages, buttons, and grading tools work smoothly without errors
Challenges we ran into
Local Code Isolation: Our project was initially saved only on one computer, making it difficult for both members of our team to share the project.
Image Input vs. Output: When we were integrating Hugging Face, generated images kept being read as input data rather than being rendered as output in the interface, which made the computer try to “eat” the pictures it generated instead of showing them on screen.
Inaccessible Local Demo: The website was running on localhost:8502, which meant people outside of the wifi we were on unable to access our entire project.
- Git & Terminal Setup Errors: We ran into terminal environment errors when attempting to run local Git commands without Git installed on our operating system because we didn’t install the tool we needed to save and send code, giving us many error messages.
Accomplishments that we're proud of
Fully Working AI Assistant: Built GradePilot into a functional tool that actually helps integrate AI into the education sector while still allowing control of teachers. Additionally, we made this tool an assistive tool for teachers, not something they must use or end up wasting more time for them in the long run.
Google AI Integration: We successfully connected AI models to grade papers and give feedback without crashing.
Live Online Web Demo: Our site is actually online, so all sorts of teachers can use our app in the real world.
Resilient Teamwork: Our team worked very well together fixing errors and finishing our project as fast as possible.
What we learned
Browser-Based Version Control: We learned how to manage GitHub repositories and commit files directly through the web interface when local Git isn't available.
Handling AI Image Data: We gained experience debugging model pipelines in Python to properly format and display AI-generated image outputs.
Cloud Deployment Workflows: We learned how to connect GitHub repositories to Streamlit Community Cloud to convert local localhost apps into public web links.
Creative Problem Solving: We were able to work around any issue that we had stuck up to us, allowing us to learn complex problem solving on a time crunch.
What's next for GradePilot
Integration: We hope to integrate our tool into the real world, allowing teachers to actually save time in their daily lives.
StudentPilot: We hope to create a tool that can rival Canvas by creating a student view area, allowing students to use AI tools to teach them within their actual assignment setting, along with seeing their grades and a helpful AI summary of how they can actually improve.
LessonPilot: We hope to create a tool that can also make powerpoints for teachers to teach exactly what needs to be taught, without the hassle of creating Powerpoints or Canvas.
Log in or sign up for Devpost to join the conversation.