Machine Learning as a Service for HEP

TL;DR

This paper proposes a Machine Learning as a Service (MLaaS) framework tailored for High-Energy Physics, enabling scalable training and deployment of ML models directly on CERN's large ROOT data format using distributed infrastructure.

Contribution

It introduces a modular MLaaS architecture that reads ROOT data, leverages WLCG for remote access, and serves models via HTTP, facilitating large-scale training and easy deployment in HEP.

Findings

Supports native ROOT data format processing

Enables remote data access via WLCG infrastructure

Provides HTTP-based model serving for integration

Abstract

Machine Learning (ML) will play significant role in success of the upcoming High-Luminosity LHC (HL-LHC) program at CERN. The unprecedented amount of data at the Exa-Byte scale to be collected by the CERN experiments in next decade will require a novel approaches to train and use ML models. In this paper we discuss Machine Learning as a Service (MLaaS) model which is capable to read HEP data in their native ROOT data format, rely on the World-Wide LHC Grid (WLCG) infrastructure for remote data access, and serve a pre-trained model via HTTP protocol. Such modular design opens up a possibility to train data at large scale by reading ROOT files from remote storages, avoiding data-transformation to flatten data formats currently used by ML frameworks, and easily access pre-trained ML models in existing infrastructure and applications.