#64 - CheatGPT | Devpost

Most Popular Questions
Up-To-Date Information
Stores Chat History
Architecture Diagram

Inspiration:

We wanted to help prospective students learn about Georgetown efficiently and effectively.

What it does:

Our virtual admissions bot helps high school students answer their common questions about going to college, such as the application process and timeline, campus life, paying for school, and the course catalog.

How we built it:

We connected the frontend directly to Azure AI (deployed GPT 4 preview), connected Azure AI to Azure CosmosDB for MongoDB, and populated MongoDB with data scrapes from major parent sitemaps of Georgetown University as of today.

Challenges we ran into:

We learned to work with the different Azure services, individually and when connecting them with each other. We were only able to deploy GPT 4 preview in Azure OpenAI Studio

Accomplishments that we're proud of:

We created an Azure-based chatbot that answers Georgetown related questions more accurately than ChatGPT4. Our chatbot also includes hyperlinks in its answers and allows users to access past chat histories.

What we learned:

We learned to use Azure services such as CosmosDB for MongoDB, Azure AI Search, and Azure OpenAI Studio and design professional UI/UX.

What's next for HoyaHelper:

Continue to build up training data as users provide feedback on topics like social experience that are less tangible & less directly available on public sites.

What's next for HoyaHelper

Improve web scraping framework to recursively search URLs and get more department–specific information. Connect Azure services to the user interface through formal server architecture.

Built With

azure
azure-ai-search
openai
python
react
rest

Submitted to

Hoya Hacks 2024 Powered by Cloudforce
- Winner 2nd Place
- Winner Cloudforce-Microsoft AI Track

Created by

I worked on connecting Azure Search Service to OpenAI and Cosmos DB for MongoDB, training our chatbot, and debugging our Python code.

Annika Lin
Senior studying CS/Econ at Georgetown University
I worked on the webscraping script, writing data to the CosmosDB and building Azure AI Search on top of it. This was my first time working with XML sitemaps, but it made webscraping very logical. It was also cool learning about how to integrate the various Azure services.

Sameer Tirumala
I worked on developing our front-end using NextJS, mocking the custom UI in Figma, and interfacing with the model on the Azure back-end to display query results to users.

Reed Uhlik
I designed and diagrammed the architecture, fixed connectivity issues between Azure Search and OpenAI, and debugged the front end by incorporating various critical parameters. I enjoyed learning Azure and look forward to exploring more in the future!

Maggie Shen