Scheduler

Inspiration

We were inspired by the countless amount of times that we procrastinated on a homework, leading to bad grades, or very rushed all-nighters. Therefore, we have decided to create a tool, that can help students better manage their time doing their hardest homeworks, in conjunction with their time constraints (lunch, football practices, study break, ...)

What it does

A user profile will include the student's past performance, skills, time constraints. Whenever a new document is uploaded, an LLM, coupled with a matching and scheduling algorithm will automatically populate the student's schedule, based on that particular homework/document. The populated schedule will take in consideration the student's time constraints and skills.

How we built it

We've built the frontend in NextJS coupled with TailwindCSS and shadcn/ui for styling and components, we have also included a very thorough authentication flow using Auth0, as the Auth0 configuration will automatically trigger an action to add new users' information, profiles to a MongoDB database. For the backend, we have decided to use a NodeJS "Express" application, with Typescript support, coupled with Intel AI PC's for local hosted LLMs. For the LLMs to accurately index and infer the best schedule for the student, we combine the high performance Vector Search capabilities of MongoDB, as well as the various embedding models provided by the Intel AI PCs.

Challenges we ran into

We've run into various problems regarding the LLM's performance in creating a suitable schedule for the user, as well as having troubles connecting and indexing from the MongoDB Vector database.

Accomplishments that we're proud of

We are proud that we were able to use Intel AI's LLM products, in conjunction with LLMs frameworks that we have utilised in the past (ollama, langchain).