Skip to content

Quickstart Guide

This guide walks you through running your first machine learning experiment and viewing the results.

Step 1: Check the Sample Data

The repository includes a ready-to-use customer churn dataset in data/raw/: - data/raw/churn_train.csv: Training dataset containing 400,000+ customer records. - data/raw/churn_test.csv: Holdout testing dataset.

Columns in this dataset include: - CustomerID: Customer identifier. - Age, Gender, Tenure: Demographics. - Usage Frequency, Support Calls, Payment Delay: Usage statistics. - Subscription Type, Contract Length, Total Spend: Account features. - Churn: The target column (1 = churned, 0 = retained).

Step 2: Inspect the Configuration

Take a look at configs/mvp.yaml. It defines: 1. Experiment Name: customer_churn_experiments 2. Data: Loads data/raw/churn_train.csv, marks Churn as target, and splits 20% for testing. 3. Preprocessing: - Renames columns to lowercase snake_case. - Casts Age, Gender, and Tenure types. - Imputes missing Age with the mean and missing Gender with the mode. - Encodes Gender and Contract Length ordinally, and Subscription Type via one-hot encoding. - Standardizes Age and min-max scales Total Spend. 4. Model: Fits a LogisticRegression classifier with max_iter: 250. 5. Evaluation: Computes accuracy and f1 score.

Step 3: Run the Experiment

Run the pipeline using main.py:

uv run python -m main --config configs/mvp.yaml --logging-level DEBUG

python -m main --config configs/mvp.yaml --logging-level DEBUG

CLI Options Explained

  • --config (required): Path to the YAML file.
  • --logging-level (optional, default: INFO):
  • Use DEBUG if you want to see detailed step-by-step diagnostic logs.
  • Use INFO for standard progress logs.
  • Use WARNING or ERROR for quiet execution.

Step 4: Read the Console Logs

When you run the command, you will see color-coded logs showing progress:

DEBUG: 2026-10-14 23:28:15 : main.py : [run] : 11 : Reading data from config
INFO : 2026-10-14 23:28:15 : orchestrator.py : [build] : 52 : Starting pipeline build for model : customer_churn_experiments
DEBUG: 2026-10-14 23:28:15 : loader.py : [_load_train_data] : 78 : Training dataset shape : (400000, 11)
INFO : 2026-10-14 23:28:15 : orchestrator.py : [build] : 71 : No test dataset provided, proceeding with train-test split.
INFO : 2026-10-14 23:28:16 : orchestrator.py : [build] : 82 : Building preprocessing pipeline...
INFO : 2026-10-14 23:28:16 : orchestrator.py : [build] : 102 : Pipeline instance created. Pipeline ready for training and testing...
INFO : 2026-10-14 23:28:18 : orchestrator.py : [train] : 121 : Training completed, pipeline fitted with training dataset.
INFO : 2026-10-14 23:28:18 : orchestrator.py : [evaluate] : 146 : --- Starting Evaluation (on provided dataset) ---
INFO : 2026-10-14 23:28:18 : orchestrator.py : [evaluate] : 158 : | Accuracy  : 0.8425
INFO : 2026-10-14 23:28:18 : orchestrator.py : [evaluate] : 158 : | F1 Score  : 0.8110
INFO : 2026-10-14 23:28:18 : orchestrator.py : [evaluate] : 160 : --- Evaluation Complete ---

Step 5: Check Your Run in MLflow

  1. Open http://localhost:5000 in your browser.
  2. Click on the experiment customer_churn_experiments in the left sidebar.
  3. You will see your run listed with:
  4. Metrics: Accuracy, F1 score.
  5. Artifacts: Look inside the logs/ folder in the MLflow UI to view the exact log file for that execution!

Step 6: Try XGBoost

To train an XGBoost model instead of Logistic Regression, simply pass the XGBoost configuration:

uv run python -m main --config configs/mvp_xgboost.yaml --logging-level INFO

Notice that you didn't have to rewrite a single line of Python code—the pipeline handled the model change automatically through configuration.