mulhamfetna/kaggle-playground-series-s6e4

Python

0

0

Kaggle Playground Series S6E4 - From Competition Model to Deployable API

#

This repository demonstrates a full path from Kaggle tabular modeling to a production-style inference service.

What this project proves

#

1. Competitive ML workflow for playground-series-s6e4 (Predicting Irrigation Need).
2. Reproducible training pipeline with configurable compute budgets.
3. Deployable serving layer (FastAPI + Docker), not just notebook experimentation.

Repository structure

#

• kaggle_gpu_submission_workflow.ipynb
  Kaggle Notebook workflow to train/evaluate and generate submission_ready.csv.

• train_advanced_and_submit.py
  Advanced competition trainer (CatBoost + CV + optional search + report JSON).

• kaggle_submission_guide.md
  Step-by-step notebook submission guide.

• train_api_model.py
  Trains a CPU CatBoost model and exports serving artifacts to model_artifacts/.

• app/main.py
  FastAPI service with:

  • GET /health
  • GET /schema
  • POST /predict

• Dockerfile + requirements.txt
  Containerized API deployment stack.

Architecture

#

flowchart LR
    A[train.csv] --> B[Feature selection + categorical preprocessing]
    B --> C[CatBoost training]
    C --> D[Model artifact .cbm]
    C --> E[Metadata JSON]
    D --> F[FastAPI inference service]
    E --> F
    F --> G[/predict -> Irrigation_Need class/]

Kaggle workflow

#

1. Open the competition notebook environment.
2. Run kaggle_gpu_submission_workflow.ipynb.
3. Produce:
   • /kaggle/working/submission_ready.csv
   • /kaggle/working/model_report.json
4. Submit in competition UI.

API training and local serving

#

1) Install dependencies

#

pip install -r requirements.txt

2) Train and export API artifacts

#

python train_api_model.py --data-dir . --output-dir model_artifacts --iterations 700 --seed 42

3) Run FastAPI

#

uvicorn app.main:app --host 0.0.0.0 --port 8000

4) Test endpoints

#

curl http://127.0.0.1:8000/health
curl http://127.0.0.1:8000/schema

Docker deployment

#

Build image (after generating model_artifacts/):

docker build -t irrigation-api:latest .

Run container:

docker run --rm -p 8000:8000 irrigation-api:latest

Business framing

#

This repo is designed as a public proof asset for ML/Data Engineering work:

• measurable leaderboard performance
• reproducible experiments
• deployable inference interface
• clear operational documentation

mulhamfetna/kaggle-playground-series-s6e4

Python

0

0