²ÝÝ®ÊÓƵ

In this blog, Connor Rogstad and Ella McMillan explain their summer work on the CATLab AI Team and the commitment that it takes to begin a years-long technology project.

This summer, we had the incredible opportunity to work with CATLab on the AI development team, where we created ²ÝÝ®ÊÓƵ.ai—a chatbot designed to assist ²ÝÝ®ÊÓƵ students with a variety of inquiries. 

As students at ²ÝÝ®ÊÓƵ, we often encountered challenges in finding information about various aspects of college life. While the ²ÝÝ®ÊÓƵ staff are amazing and always willing to help, they are often overwhelmed with the volume of student questions. Our goal was to lighten this load by creating a tool that could provide quick and accurate responses to student questions, thus freeing up staff to focus on other issues. Our project aimed to streamline access to information about academics, campus life, college policies, and more.

²ÝÝ®ÊÓƵ.ai, inspired by a similar project at Harvard, involves three main interactions: between the user and the chatbot, the chatbot and the database, and the chatbot and OpenAI's LLM. Here's a breakdown of how it works:

1. User Interaction: The user accesses ²ÝÝ®ÊÓƵ.ai via a secure login page, ensuring that only those with proper ²ÝÝ®ÊÓƵ credentials can use the service. Once logged in, users can ask questions through the chatbot interface.
2. Data Processing and Embedding: When a query is received, ²ÝÝ®ÊÓƵ.ai searches through pre-embedded documents from ²ÝÝ®ÊÓƵ’s database, selecting those that are most relevant based on a relevance score. This process ensures that the responses have contextual grounding and are based on ²ÝÝ®ÊÓƵ data.
3. Prompt ²ÝÝ®ÊÓƵ and Refinement: The chatbot uses a pre-engineered prompt template, which includes the user's question and relevant data, to generate a response from the LLM. This structured approach ensures that the chatbot's answers are both accurate and relevant to the ²ÝÝ®ÊÓƵ.ai use case.

We deployed ²ÝÝ®ÊÓƵ.ai to AWS Lightsail, making it accessible to the ²ÝÝ®ÊÓƵ community. The estimated cost per user for maintaining this service is approximately $6.50 per academic year, a calculation based on expected usage patterns and data update frequencies.

Our work this summer laid a strong foundation for ²ÝÝ®ÊÓƵ.ai, but we see this as just the beginning and are committed to adding more features. Future enhancements will include integrating the chatbot directly into the ²ÝÝ®ÊÓƵ website or app for easier access. Additionally, we seek to incorporate additional data sources, such as My²ÝÝ®ÊÓƵ and the Dallas Willard database, to provide a more extensive and personalized experience. We are also pursuing strategies such as rate limiting to steward our resources effectively.

²ÝÝ®ÊÓƵ.ai is designed to be a valuable resource for all students, particularly those who may face greater challenges in navigating college life. By providing a reliable and easy-to-access source of information, we hope to bridge the information gap and support all students in their academic and personal growth. We are incredibly grateful for the support and guidance from the CATLab team and look forward to seeing how ²ÝÝ®ÊÓƵ.ai continues to evolve and benefit the ²ÝÝ®ÊÓƵ community.