AI POWERED PERSONALISED LEARNING PATH FOR STUDENTS

Overview

This project aims to develop an AI-powered system for creating personalized learning paths for students. Leveraging user-defined attributes, this system generates tailored study plans that include workshops, tasks, activities, and mentorship recommendations.

---

Implementation Details

1. LLM Integration:

   • The entire solution is powered by OpenAI's GPT-4o model.
   • The generated dataset is used to finetune "session parameters" for instances created during production use. This ensures accurate, reliable and contextually appropriate answers.
   • Explicitly passed System Prompts which help increase accuracy of the task without overhead of computation resources needed for finetuning model parameters.

2. Mock Dataset Generation:

   • Generated mock dataset with 7 student centric attributes.

     • Strengths
     • Weaknesses
     • Interests
     • Learning Style
     • Learning Challenges
     • Goals
     • Availability

   • All the attibutes represent a key area that shape the learning plan. Clever prompt engineering ensures learning paths generated suit the needs of the students.

3. Unstructured Data Handling:

   • JSON parsing and processing for creation of labelled (student_attribute, study_plan) labels for finetuning "session parameters".

4. Web Interface:

   • Developed using StreamLit to git a GUI for the student to interact with the model.
   • A single-session chatbot interface where users can input attributes and query the generated plan.

---

How to Run?

1. Clone the Repository:

   git clone https://github.com/coder-utkarshchaudhary/AI-powered-personalised-learning-path-for-students.git
   cd AI-powered-personalised-learning-path-for-students

2. Install Dependencies: Ensure pip is installed, then run:

   pip install -r requirements.txt

3. Set Environment Variables: Create a .env file in the project root and add your OpenAI API key:

   OPENAI_API_KEY=your_openai_api_key
   

4. Run the Application: Start the Streamlit application:

   streamlit run main.py

---

Final Submission

• Codebase: Link to GitHub repository
• Video: Click on image(s) below for the video demo
  
  <a href="https://drive.google.com/file/d/13izGcUk4rhv2NAgRUMfvRE8oTyFhvtSb/view?usp=sharing", target="_blank">

---

Notes

1. The project is powered by GPT-4o model. The model's API is access via On-Demand platform. The code will not directly work for OpenAI API calls directly. Please modify the get_llm_response function in utils.py.

   Blockers and Possible Fixes:

   1. LLMs can't access and obtain information about user profiles from social media websites. Hence a lot of entires for mentors in the dataset are "John Doe" or "Jane Doe" or "Emily ...". This can be easily fixed by developing a vector database of mentor profiles and establishing a RAG on GPT-4o and said database.

   Future Scope:

   1. Implementation of RAG and Agents on a custom dataset (scapped/collected).
   2. Validation of results via feedback loop of multi LLM agentic framework.
   3. Multi-chat support on the UI.

---

Acknowledgement

Thanks to the open-source libraries and tools used in this project --> OpenAI, Streamlit and On-Demand.
This project was made as a part of the Alcovia Intern Mandatory Task. I extend a heartfelt thank you to the Alcovia team for providing me this opportunity.

---

License

This project is licensed under the MIT License. You are free to use, modify, and distribute this software, provided proper attribution is given.

Follow Up Project

Also check out the extension of the project at AI-powered-mentor-mentee-matching-system