UManitoba AI: Smart Campus Chat with Personalized Student Insights

Inspiration

As a student at the University of Manitoba, I noticed that finding accurate and specific information from the university’s website could be time-consuming. Additionally, students often need personalized data related to their academic records, course schedules, and university policies. This inspired me to build an AI-powered chatbot that not only retrieves general university information but also integrates with a student database to provide personalized insights.

What I Learned

Throughout this project, I gained valuable experience in:

• Web Scraping & Data Processing: Extracting structured data from the university’s website using Scrapy and Selenium.
  
• Fine-Tuning AI Models: Training a custom GPT model on the university's data to improve accuracy and relevancy.
  
• Database Management: Integrating a student database for personalized responses.
  
• Flask & API Development: Building a backend server to handle user queries and AI responses.
  

How I Built It

1. Data Collection: Used Scrapy and Selenium to gather text-based information from the University of Manitoba’s website.
   
2. AI Model Fine-Tuning: Processed the scraped data and fine-tuned a Hugging Face GPT model with domain-specific knowledge.
   
3. Backend Development: Created a Flask API to handle chat requests and fetch personalized student data from a database.
   
4. Database Integration: Connected the chatbot with a student database to retrieve course schedules, grades, and other relevant details.
   

Challenges Faced

• Model Accuracy: Ensuring the chatbot provided relevant and factually correct responses. I was able to just train 1 GB data within the time limit.
• Data Cleaning: The university website contained unstructured data, requiring extensive preprocessing.
  
• Database Security: Implementing proper authentication and access control for sensitive student data.
  
• Latency Issues: Optimizing API response times to make the chatbot feel responsive.
  

Future Improvements

• Voice Integration: Adding voice-based interaction for accessibility.
  
• Multilingual Support: Expanding the chatbot to support multiple languages.
  
• Improved Personalization: Enhancing AI-driven recommendations based on student queries.
  
• Student profile analysis and Recommending Courses:If tarinned and connected to university databases if can give personlized response and analysis. Enhancing AI-driven recommendations based on student queries.
  

This project was an exciting journey that combined AI, web scraping, database management to create a helpful tool for students. 🚀

Built With

• python
• sqlite