Machine Learning Operations: From a Notebook to a Live Model

by Amine Lahouel (CERN), Raulian-Ionut Chiorescu (CERN)

Friday 24 Jul 2026, 14:00 → 16:00 Europe/Zurich

593/R-010 - Salle 11 (CERN)

Abstract

A machine learning model is only useful when other people can run it. This session shows the full Machine Learning Lifecycle path from an idea to a model that answers requests in production.

 

We first explain the machine learning workflow: data preparation, training, tuning, serving and monitoring. We then show the CERN platform that supports it, providing seamless access to hardware, curated software environments and shared storage.

 

In the hands-on part, each participant trains a model on a real GPU. You then build a pipeline, scale the training across several workers, tune the hyperparameters automatically, and deploy the model as a live inference service. Finally, you send requests to your own model and watch it scale under load.

Lecturer Bio

Amine Lahouel is a Software Engineer in the CERN IT department. He works on the Scientific Computing platform of NGT, making MLOps accessible to everyone. Before joining CERN in 2025, he participated in building various platforms, from full-stack apps to web-scale fast crawlers and ML serving platforms for search engines.

CERN STEAM Academy 2026 - Machine Learning Operations.pdf