Aggie Parts Picker

Inspiration

As freshmen, we are often asked a lot of ourselves, meeting people, getting acclimated to our new amazing home, leaving meal planning and cooking from scratch a much less desirable option despite its inherent potential to be much more healthy and enjoyable than overly repetitive school meal plans and expensive takeout.

What it does

We use a database that’s automatically populated with recipe information, alongside its generated embedding, from web scraping websites in order to use RAG to create an AI model that can intuitively respond to text queries from the users about recipes of interest. The AI model then outputs recipe information that can be sorted into meal plans (cook at home or restaurants), which are then sorted into the user’s school schedule, adjusting for travel time and distance using Google Maps API.

How we built it

After using python and mySQL to create a database of recipes and embedding information, we used AWS bedrock to create a model that uses the database to dynamically respond to user queries about recipes they wanted to make. Additionally, we created microservices in Rust to interface scheduling, data management, and other services. Lastly, we have a frontend application built on the backs of nextJS and React.

Challenges we ran into

From the start, we had issues with scaling due to the amount of features we planned to implement into the application. We were very ambitious with our designs as we felt that this had the potential to greatly impact the Aggie community, especially for freshmen with packed schedules and tumultuous starts.

Accomplishments that we're proud of

Robust datasets generated from data scraping several websites, full stack web design with complete backend and front end, integration of google maps API into meal planning, automated database data population and embedding generation, using RAG to create insightful responses and harness the power of AI.

What we learned

It is important to manage the scope of our project and keep it scaled to a specified time frame, as well as make end-to-end design plans and scouting out several alternative solutions before deciding on what frameworks to use, as well as making more tactical software architectural solutions.

What's next for Aggie Parts Picker

We plan to fully connect the remaining features into a cohesive workflow for the users on our application, and hope to release our application to the Aggie community. We also have more features planned out of the future: We hope to reach a critical mass of users such that our machine learning system learns to load balance students when routing students between and to locations. Additionally, we hope to create a comprehensive realtime repository of input prices for food and other commodities and use this data in addition with student finances to give them an optimal report on how they should partition their money in regards to food, commodity, and time consumption.