Adaptive Coding Tutor using LLM

Given a student’s coding activities: how they wrote each line of code, the kind of coding errors, whether the code was optimized or not, etc, generate personalized lessons and practice exercises.

Comment

Inspiration

In a programming classroom, not all students learn coding at the same pace. Some are quick learners and some are slow learners.

What it does

The app generates personalized lessons and practice exercises for a student after assessing their coding skills and activities: how they wrote each line of code, the kind of coding errors, whether the code was optimized or not, etc.

How we built it

Angular + FastAPI + Docker + LLM + Python + ChatGPT

Challenges we ran into

I was not able to access LLM on AWS through Bedrock till the submission deadline because of an account issue for which the root cause was not known. Update: Switching to ChatGPT at lunchtime. Was able to integrate and generate coded review and use it to generate the suggested lesson plans.

Accomplishments that we're proud of

  1. I was a solo person working on this idea.
  2. I wrote a dockerized backend REST service utility to compile and run a python script send by a client through a POST endpoint. This acts like a python code compiler API.
  3. I wrote a code editor from scratch using Angular and TypeScript

What we learned

Working on the front end using the latest Angular was a humbling experience and realized that my previous knowledge of the older version was no longer compatible with newer libraries to quickly spin up full-featured code editor.

What's next for Adaptive Coding Tutor using LLM

Experiment the product with real consumers (students at the University of California, Merced) and do usability testing.

Built With

Share this project:

Updates