• 

  Ask HOKIE

• 

  Problem Statement

• 

  Technical Framework

• 

  Database Flowchart

• 

  Impact

• 

  Testimonials

Motivation:

The inspiration for developing a chatbot for Virginia Tech came from the desire to enhance the student experience by providing a readily accessible, intelligent assistant. With increasing demands for real-time information and support, we aimed to create a tool that would simplify and streamline access to campus resources, answer frequently asked questions, and offer personalized assistance to students, faculty, and visitors.

What it does:

The Virginia Tech chatbot serves as a virtual assistant designed to: •⁠ ⁠Provide instant answers to common questions related to admissions, campus facilities, academic programs, and events. •⁠ ⁠Assist with navigation around campus and offer information about departmental services. •⁠ ⁠Facilitate student engagement by offering details about extracurricular activities, clubs, and support services. •⁠ ⁠Provided responses for emergency response in case of natural calamities and criminal activities.

How we built it:

Our platform uses Python libraries (openai, pymongo, fastapi, sentence transformers, pydantic, pytorch, requests, and beautifulsoup), MongoDB Atlas, GPT-4-Turbo, PropelAuth, and React.

Challenges We Ran Into:

We think the biggest challenge was data collection as Virginia Tech's website has a lot of information, and collecting and processing takes a lot of time. We also faced challenges in integrating multiple modules together.

Accomplishments We're Proud Of:

Even after the above-mentioned challenges, we were successfully able to accomplish most of our goals. The biggest accomplishment for us is that we worked with LLMs for the first time and were able to create a good chatbot that is able to answer most questions.

What We Learned:

We were able to learn a lot of skills over this hackathon. Interacting with other hackers and industry experts we were able to gain a lot knowledge. We attended various workshops and incorporated the methods taught. We learned about LLMs and how RAG can be used to refine the LLM results.

What's Next for Ask Hokie

We aim to include other VT domains like offcampus.vt.edu and cs.vt.edu. These would help us to make the chatbot more concise and integrated.

Built With

• css3
• gpt
• llm
• mlh
• mongodb
• propelauth
• react