← All projects

MoleCrab MLOps Platform

Project description

A Kubernetes-based MLOps project for reproducible image-model training, evaluation, registration, serving, and monitoring. Its architecture separates infrastructure, GitOps, ML workflows, model lifecycle management, inference, and observability.

Macro diagrams

High-level views of the system, MLOps lifecycle, training pipeline, and model serving.

01 · System architecture

System Architecture

High-level view of the offline training system, online prediction system, GitOps, and infrastructure.

02 · MLOps lifecycle

End-to-End MLOps Flow

High-level flow from versioned data and training through model approval, deployment, serving, and monitoring.

03 · Model training

Training Pipeline

High-level training workflow covering preparation, validation, experiments, evaluation, and model registration.

Diagram