Diamonds — ML Production Challenges

This repository is a learning sandbox for ML engineering and MLOps practices.

The course is split into 5 challenges, and each challenge is delivered on its own Git branch. This README stays generic and gives only the common workflow used across all branches.

Repository Structure

• src/diamonds/: Python package code
• data/: raw and preprocessed data
• models/: locally saved models
• mlartifacts/: MLflow artifacts
• notebooks/: exploration and experimentation notebooks
• docs/: challenge instructions

Current docs already available:

• docs/01-Packaging.md
• docs/02-Model-Registry.md

Challenge Model (5 Branches)

At the end, the repository contains 5 challenge branches (one branch per challenge).

Typical flow:

1. Checkout the branch for the challenge.
2. Read the matching file in docs/.
3. Implement the requested tasks.
4. Commit and push your solution branch.

Quick Start (Common to All Challenges)

From the project root:

make setup

This installs the project in editable mode (pip install -e .).

Python requirement (from pyproject.toml):

• Python >=3.11,<4.0

Branch Workflow

git fetch origin
git checkout <challenge-branch>
git pull origin <challenge-branch>

If you work on your own solution branch:

git checkout -b <your-name>-<challenge>

MLflow (for tracking/registry challenges)

Start MLflow server:

make launch-mlflow-server

Default local UI:

• http://127.0.0.1:5000

Optional environment variable for scripts:

export MLFLOW_TRACKING_URI=http://127.0.0.1:5000

Notes

• Keep changes focused on the current challenge branch.
• Do not assume code from another challenge branch is already merged.
• Prefer small, clear commits with explicit messages.