Clarifii

Website Homepage
Upload Page
Adaptive Quiz

Inspiration

As CSC TAs, we've had to help our fair share of confused and frustrated Computer Science students.

We wanted to build a tool that would both help us in conducting our TA jobs and in preparing for our exams in the future.

We hope Clarifii can be the needed breakthrough that students and TAs can use to advance their education

What it does

Clarifii serves 3 main functions:

Generate Interactive lesson explanations that utilize various attenion grabbers to keep students engaged.
Take in homework .pdf files and generate adaptive quizzes that equip students with the skills and knowledge needed to complete the homework provided.
Create exam style questions to have extra material to practice for exams.

How we built it

We used Django for our server framework and the main programming language we used was python.
Implemented the Large Language Model from MLC Engine paired with WebGPU to allow for querying in-browser

Challenges we ran into

Hosting the open source LLM in a fashion such that we could serve accurate and safe responses.
We had issues merging mp3 and mp4 files without corrupting and making sure the captions, audio, and background overlay synced up with each other.

Accomplishments that we're proud of

Completing pythonic video editing with 'Moviepy' that merges video, audio and subtitle text in real time.
Splicing and filtering mp3 and mp4 files for merge compatibility
Designing a simple, yet drawing landing page for our Web app

What we learned

Video manipulation with moviepy and mlcEngine using webGPU.

What's next for Clarifii

async server side code has been written but not fully debugged. Upon completion we hope to acquire a large enough compute pool to run Clarifii smoothly for up to 10,000 users continuously.