Experimentation, deployment, and monitoring of machine learning models - How MLOps enhances AI productization

Aug 3, 2026

Data science increasingly supports industrial decision-making, while companies still struggle to move machine learning models into productive environments. Manual processes often introduce errors, reduce reproducibility, and waste effort, limiting scalability and delivery speed. MLOps automates the end-to-end lifecycle from experimentation to deployment and monitoring, improving governance and operational reliability. The paper presents three successful case studies, including an application in a major Latin American financial bank, to show practical value of an MLOps pipeline implementation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Addressing Key Model Lifecycle Challenges with MLOps

Association for Information Systems AIS Electronic Library (AISeL) | ISLA 2025 Proceedings | Latin America (ISLA) | | --- | --- | | Winter 12-31-2025
Experimentation, deployment, and monitoring of machine learning models: How MLOps enhances AI productization
Diego Nogare
Guilherme Henrique Iglesia da Silva Ismar Frango Silveira
Follow this and additional works at: [https://aisel.aisnet.org/isla2025](https://aisel.aisnet.org/isla2025) | | This material is brought to you by the Latin America (ISLA) at AIS Electronic Library (AISeL) . It has been accepted for inclusion in ISLA 2025 Proceedings by an authorized administrator of AIS Electronic Library (AISeL) . For more information, please contact [elibrary@aisnet.org](elibrary@aisnet.org). How MLOps enhancesAI productization Experimentation, deployment, and monitoring of machine learning models: How MLOps enhances AI productization Diego Nogare Universidade Presbiteriana Mackenzie -PPGEEC [diego.nogare@gmail.com](diego.nogare@gmail.com) Guilherme Henrique Iglesia da Silva Itaú Unibanco [guilhermeiglesiaspro@gmail.com](guilhermeiglesiaspro@gmail.com) Ismar Frango Silveira Universidade Presbiteriana Mackenzie-PPGCA[ismar.frango@gmail.com](ismar.frango@gmail.com) Abstract In recent years, Data Science has become increasingly relevant as a support tool for industry, profoundly impacting decision-making. With the advent of new technologies and tools, companies have begun to recognize the power of AI in addressing everyday use cases, especially in this decade. However, despite its relevance, many companies face significant obstacles to drive Machine Learning models to productive environments, and performing some processes manually can lead to errors, low reproducibility, and inefficiency. In this context, the MLOps discipline emerges as a solution to automate the life cycle of Machine Learning models, ranging from experimentation to monitoring in productive environments. This work explores how MLOps pipeline implementation can help mitigate or even eliminate many of these challenges by presenting three successful case studies, including notable application in the largest financial bank in Latin America. Keywords: MLOps; Model Experimentation; Model Deployment; Model Monitoring 1. Introduction For more than a decade, Data Science has transitioned from a niche research area to a fundamental discipline and a support tool for the industry, enhancing its data-driven decision-making. With the increasing relevance of this area, the challenges of deploying the developed models into production to deliver the proposed value to end-users have become more prominent. Despite global challenges, as presented in section 3.2 with three worldwide use cases, the Latin American region has worked diligently and made significant progress. To address these challenges, the MLOps (Machine Learning Operations) discipline has proven to be an emerging approach, enabling the automation and governance of the processes of experimenting, deployment and monitoring Machine Learning models. The creation and adoption of MLOps pipelines represent key strategies to ensure the effectiveness, scalability, and efficiency of these processes. Information Systems in LatinAmerica (ISLA 2025) 1 How MLOps enhancesAI productization To illustrate this, the present paper brings a use case that demonstrates how a financial institution benefited from the automation of data science model deployment by implementing MLOps pipelines, driving innovation and efficiency, which are important features especially in the Brazilian context. By integrating development and operational practices, MLOps can foster significant improvements and innovations in AI development workflows, encouraging the use of more robust, efficient, and transparent methodologies for end-to-end Machine Learning projects. From initial experimentation to deployment and continuous monitoring of models in production, MLOps

FAQs

What problem does MLOps aim to solve for machine learning teams?

MLOps addresses difficulties in moving machine learning models into productive environments by automating the model lifecycle. It reduces manual errors, improves reproducibility, and increases efficiency from experimentation to monitoring.

Which stages of the ML lifecycle are covered by an MLOps pipeline?

An MLOps pipeline supports experimentation, deployment, and continuous monitoring of models in production. It provides automation and consistency across all stages.

How does the paper validate the impact of MLOps?

The work presents three successful case studies, including a notable application in the largest financial bank in Latin America. These cases illustrate improvements in automation, innovation, and efficiency.

Aug 3, 2026

Uploaded by Seraphina