Python Mono-repository example

This is a simple example of a mono-repository for a typical Data Science project. This project uses uv for package and workspace management.

The project is composed of multiple components:

Libraries

The packages directory contains the stand-alone libraries used in the project.

fraud-ml

Sample library containing model architectures, training, and evaluation code.

data-connector

Sample library containing tools to connect to external data sources.

Workflows

The workflows directory contains the workflows used in the project

batch-proc

Sample project containing runnable notebooks for batch training or batch inference locally or inside a cluster. It uses the fraud-ml library for model training and evaluation, and the data-connector library for fetching historical data.

  +-------------+
  | fraud-ml |
  +-------------+
         \
          \
           \
  +------------------+
  |   batch-proc  |
  +------------------+
           /
          /
         /
  +-----------------+
  | data-connector |
  +-----------------+

Experiments

The experiments directory contains the experiments used in the project, e.g. testing out different model architectures.

  +-------------+
  | fraud-ml |
  +-------------+
         \
          \
           \
  +------------------+
  |   experiments  |
  +------------------+
           /
          /
         /
  +-----------------+
  | data-connector |
  +-----------------+

Application/Service

The app directory contains a FastAPI service for serving the model and exposing an API. It uses the fraud-ml library for model inference, and the data-connector library for model loading.

  +-------------+
  | fraud-ml |
  +-------------+
         \
          \
           \
  +-----------------------+
  |   fraud-detect app  |
  +-----------------------+
           /
          /
         /
  +-----------------+
  | data-connector |
  +-----------------+