Inspiration
Many students turn to LLMs for help understanding complex concepts and debugging code. However, a typical AI assistant often provide solutions that are too complete, enabling students to copy-paste the answer rather than truly learning and understanding topics. We wanted to create an AI tutoring system that is specifically designed for computer science education. A model that guides students through the learning process rather than simply handing out the answer.
What it does
The CSTutorbot is a itelligent tutoring platform that helps computer science students by providing contextual guidance that is tailored towards their specific courses. Students can select their class from the front-end interface, which will then contect them to a course-specific AI assistant that is powered by AWS. In which it is designed to provide hints, explanations and conceptual guidance rather than generating complete code solutions. Therefore encouraging students to have a genuine understanding of course concepts and still having academic integrity.
How we built it
We developed CSTutorbot using several different technologies:
- Built the front-end interface in VSCode, creating an intutive class selection portal
- Deployed and hosted LLM infrustructure on AWS
- Developed a secondary web interfface that commincated with the AWS hosted LLM
- Integrated both platfoms to enable seamless data flow between the selection system for each class and the AI tutoring backend -Used GitHub for collaborative development and version control
Challenges we ran into
- LLM Website Intergration: EStablishing a reliable connextion between our web interface and the language model require navigating complex configerations and protocols
- AWS Depolyment: Learning AWS infastructure and propertly configuraing services to host and serve our LLM as beginners was quite a diffcult learning curve.
- Full-Stack DEvelopment: Building a cohesive user experience that seamlessly connected the front-end class selection interfaces made with the AWS-powered backend presented significant challenges
- Response Filtering: Creating and designing prompt engineering strategies to ensure that the LLM provides educational guidance rather then the typical direct answer
Accomplishments that we're proud of
- Successfully created a front end portal where students can select their course
- Deployed a fully functional LLM system on AWS that can handle real-time student question and provided answers
- Acheived a seamless integration between the web platforms, creating a smooth user experience
What we learned
- How to create and deploy clous based AI applications using AWS services
- How to Intergrate an LLM into a web application through API management
- Full stack devlopment techniques for connecting multiple web interfaces
- Important of responsible AI design in educational contexts, how to balance helpful responses that promote genuine learning and just giving the full answer to students
What's next for CSTutorbot
- Enhanced Academic Intergrity Checks: Implement a pre-process layer that analyzes a student's query for potential code of condict violations before generating a response, to ensure that the plaform promotes learning rather than academic dishonesty
- Refined User Interfaces: Redesign portions of the web application's layout for improved usability, includung better response formatting, more intuitive navagiation between different sections of the platform, etc.
- Course-specific AI Models: Develop specialized LLM configurations that adapt their knowledge base, teaching style and response patterns based on the selected course, ensuring that students recieve accurate, cirrcumlum-alligned guidance that is tailored to each class's specific content and learning objectives
- Expanded Course Catalog: Add support for addition computer science courses and potentially different subjects
Log in or sign up for Devpost to join the conversation.