Purdue Schedulings

• 

  Welcome screen

• 

  Select your major

• 

  User interface

Inspiration

Boiler Schedulings is an innovative web application designed to revolutionize the way students plan their academic schedules. Using the latest AI technology, including a Retrieval Augmented Generation (RAG) system and advanced vector databases, our application provides personalized, efficient, and interactive scheduling assistance.

What it does

At its core, Boiler Schedulings is more than just a scheduling tool; it's an intelligent assistant that understands the unique needs and preferences of each student. Boiler Schedulings utilizes RAG over the Purdue course catalog and Purdue teacher ratings and average GPA data to provide intelligent and accurate advice to users.

How we built it

The application's intelligence extends to its use of vector databases, enabling it to handle and process large amounts of data with remarkable speed and accuracy. This allows for real-time schedule optimization, taking into account factors such as course load, professor evaluations, and class times to create the most effective schedule tailored to each student's academic goals and lifestyle.

Challenges we faced

Developing Boiler Schedulings presented significant challenges, particularly in integrating the Retrieval Augmented Generation (RAG) system with vector databases to balance AI response accuracy with real-time interaction. Creating an intuitive yet sophisticated user interface, especially for displaying dynamic widgets, required careful design and front-end development. In addition, managing asynchronous state updates in React presented challenges to ensure real-time data reflection and optimal performance. These hurdles not only increased the robustness of the application but also greatly enhanced our team's technical expertise and problem-solving skills.

Challenges we faced

Developing Boiler Schedulings presented significant challenges, particularly in integrating the Robust Answer Generation (RAG) system with vector databases to balance AI response accuracy with real-time interaction. Creating an intuitive yet sophisticated user interface, especially for displaying dynamic widgets, required careful design and front-end development. In addition, managing asynchronous state updates in React presented challenges to ensure real-time data reflection and optimal performance. These hurdles not only increased the robustness of the application but also greatly enhanced our team's technical expertise and problem-solving skills. We also faced difficulties integrating the different parts of our application as we developed the front end and the back end, but through this challenge, we were able to get hands-on experience with Flask, Cloudflare workers, and Firebase.

Achievements we're proud of

Boiler Schedulings is unique in its ability to generate dynamic "widgets" - visual tools that range from class recommendations and professor ratings to a comprehensive view of semester schedules. These widgets give students an engaging and interactive way to visualize and interact with their schedules, making scheduling more than a task - it's an experience.

What we learned

We learned that custom RAG solutions involve scraping a large amount of data, which can be tedious especially if that data is locked away in difficult-to-parse formats like PDFs. We also learned from the challenges of integrating different parts of an ML-powered application, especially in terms of a rag workflow. At first, we were going to have a three-part architecture: namely a vector DB backend, a mid-layer to augment the user prompt and assemble data from various sources into the prompt, and then a frontend. We ended up settling on a more traditional front-end and back-end architecture, combining the vector DB and mid-layer.

Built With

• auth
• chromadb
• firebase
• flask
• gemini
• huggingface
• node.js
• python
• rag
• react