Imagine pushing a single commit to GitHub and automatically triggering the full lifecycle of a machine learning application from a dataset to a fully deployed, accessible service running in Kubernetes.

In this talk, we explore an end-to-end MLOps workflow that integrates FastAPI as the backend inference service and Streamlit as the user web interface. The pipeline leverages CI/CD automation to build, test, containerize, and deploy the ML model using GitOps principles using GitHub Actions and Flux.

Once deployed, users can access the machine learning application remotely, interacting with live predictions through a clean, browser-based interface. The presentation highlights how continuous integration and reconciliation processes keep the system operational, even after failures or infrastructure changes.

The key objective is to demonstrate how to measure and minimize recovery time in case of a technical incident, showing how automation ensures resilience and fast redeployment without manual intervention.

This session combines machine learning, DevOps, and cloud-native practices to showcase how reproducible, fault-tolerant ML platforms can be built entirely with open-source tools.