Inspiration
As Berkeley students, our main classes are always plagued with long office hour queues and less-than-desirable response times. We understand that the student population of our classes is simply too large for course staff to handle and increasing manpower would cost too much. So, we built GradeGuardian as a possible solution to address this problem.
What it does
Students can use GradeGuardian to ask for clarification on certain topics that were covered in the class or to review said topics in order to study. They can also utilize GradeGuardian to get practice questions from previous exams or homework as a form of practice.
How we built it
We leveraged Amazon Bedrock to power GradeGuardian using Claude 3 Sonnet and utilized the Guardrails feature to prevent unintended negative responses. We also utilized knowledge bases through S3 in order to perform RAG on class resources. The UI was powered by Gradio and LaunchDarkly's feature flags are used to quickly switch between Claude models and RAG search methods.
Accomplishments that we're proud of
Although we didn't have enough time to implement everything we wanted, we're proud of making a product that works well to a decent extent.
What's next for GradeGuardian
We plan to continue updating GradeGuardian, until it becomes a fully functional product that we can be fully proud of.
Built With
- bedrock
- gradio
- llama-index
- python
- s3
Log in or sign up for Devpost to join the conversation.