FICO Educational Challenge 2026

This repository contains the work for the FICO Educational Challenge 2026, focused on building a machine learning system for intent classification in customer collections conversations using synthetic data.

The project follows an end-to-end ML lifecycle: data generation, model training and evaluation, and final implementation and optimization.

Getting Started (For Collaborators)

Follow these steps to set up the project and start contributing.

1. Clone the Repository

git clone https://github.com/rajilsaj/FICOchallenge.git
cd FICOchallenge

2. Google Colab & Drive Setup

The notebooks are designed to run in Google Colab and require a specific folder structure in your Google Drive to load and save data.

Important: Once you open a notebook in Colab from the links below, you must go to File > Save a copy in Drive to save it to your own account before running it.

A. Google Drive Structure

Create the following folders in your Google Drive (My Drive):

My Drive/
└── FICO Analytic Challenge/
    ├── Week_05/
    ├── Week_04/
    ├── Week_03/
    ├── Week_02/
    ├── Week_01/
    ├── Model/
    └── Data/

B. Opening Notebooks in Colab

You can open the notebooks directly from GitHub using the links below:

Notebook	Link
Week 3: Synthetic Data
Week 5: Data Preparation
Week 5: BERT Fine-tuning
Week 5: Qwen Fine-tuning

3. Running & Testing

To test the notebooks on your side:

1. Mount Google Drive: Every notebook starts with a cell to mount your Google Drive. Follow the authentication prompt.
2. Verify Paths: Ensure the paths in the "Import Libraries and Set up Folder Paths" section of each notebook match your Drive structure.
3. Run All Cells: Use Runtime > Run all in Google Colab.
4. Output Verification:
   • Week 3: Check Data/ for generated synthetic conversations.
   • Week 5 (Prep): Check Data/ for _train, _test, and _validation splits.
   • Week 5 (Training): Check the Model/ folder for saved weights and evaluation logs.