Advance.io, a Group Elephant business, implements Machine Learning Operations (MLOps) and defines associated governance structures in support of MTN’s Artificial Intelligence (AI) activities.

The largest mobile network operator in Africa, MTN serves 262 million subscribers across 21 countries offering voice, data and digital services. Leveraging its extensive data footprint, MTN has wholeheartedly embraced AI, leading to new revenue streams and improved financial performance. Among MTN’s AI applications are those already enabling RICA compliance testing, user recommendations and client campaigns.

With a 15-year track record in serving MTN on several aspects of the technology front, Advance.io has played a pivotal role in implementing MTN's AI strategy. Despite employing in-house Data Scientists proficient in developing and fine-tuning machine learning models on non-production environments, MTN faced challenges in deploying and validating these models sustainably in a real-world production environment. This challenge is commonly addressed in the AI field through MLOps, the process of automating and simplifying machine learning workflows and deployments.

Leveraging a wealth of experience with DevOps best practices and IT architectures, Advance.io implemented infrastructure and processes for MTN, to allow for secure and consistent deployment of machine learning models. Docker, Kubernetes, Gitlab and PySpark were applied in a distributed fashion to satisfactorily handle the prodigious volume of MTN data.

MTN relied on Advance.io AI engineers to put in place the necessary governance, to guide Data Scientists when developing machine learning models. These guidelines are designed to allow models to be deployed to production automatically, in compliance with local privacy regulations. Governance was then fortified through minimum requirements on documentation, model training, detection of model drift and retraining. MTN now regards the resultant security, compliance and governance environment to be a gold standard for internal use, to be adopted by other MTN operating companies in due course.

Advance.io played a crucial role in transitioning MTN from experimental proof-of-concepts to the establishment of a reliable and secure production environment, where models can be deployed continuously. This transformation enables data scientists to focus on their ML models and the data, freeing them from the burden of maintaining a production environment. MTN can now measure the effectiveness of ML models, retaining only those aligned with the objectives of their AI initiatives.