Skip to content

project for ai samasya hackathon. using a novel prompt eval based prompt optimizer methodology to capture the essence of students personality and preferences continually

Notifications You must be signed in to change notification settings

Govind-S-B/adapt-learner

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

39 Commits
 
 
 
 
 
 

Repository files navigation

adapt-learner

project for ai samasya hackathon. using a novel eval based prompt/policy optimizer methodology to capture the essence of students personality and preferences continually in order to assist them with their learning and adapt existing, such as using familar sentence structures or preffered writing styles, phrasing, emotional sensitivity, relatable analogies, etc - adaptable and personalized. All this without using a standard finetuning mechanism.

TLDR ; an AI course material assisting copilot that knows about you and learns more without any heavy computation involved making it more transparent and flexible.

How to run it

cd backend
python main.py
cd frontend
npm run dev

Primary reasearch is around o1 series of models and policy optimization using RLHF methods https://vimeo.com/1023317525/be082a1029 https://vimeo.com/1018737829/ce8ca37ae2
https://arxiv.org/abs/2411.10109

image image

Original Problem Statement

How can AI driven solutions be designed to enhance personalzied learning experiences for differently able individuals ensuring accessibility, inclusivit, and adaptabe across driverse needs.

Approach Pitch

Standardized Education cannot cater to all folks. They always have to sacrifice the top and bottom 5% in this system. Some are intrinsically intuitive and like to think from first principles with challenges, some are better at memorization of facts and learn better when presented that way, some work with analogies and the worlds best teachers are good at catering to this individual need of being presented the same information in different ways.

We have a large collection of materials that was hard to adapt and it was the teacher who had to build a mental model of the student and repackage it and assist them while learning any course material.

But this is not scalable and doable by a lot of tutors to spend time and give attention to all the individual students and keep learning and tracking them.

To make education accessible and work at scale its important to solve this. So we are tackling this understanding the students habbits from feedback collected from people around them such as parents, teachers and how they react with adapted materials in a continous manner (which can also be simulated after a point) and thereby building a persona - a mental model of the student similar to how teacher have it. Now students can navigate course materials on their own and use this persona to adapt materials for their needs w/ any additional requests they have. This could also be used by parents or tutors to help guide their kids instead. Its a copilot rather than a uncontrolled entity without supervision.

Numbers and Statistics

About

project for ai samasya hackathon. using a novel prompt eval based prompt optimizer methodology to capture the essence of students personality and preferences continually

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published