MLOps Implant
Vodafone stands out globally in the telecommunications sector, backed by an advanced internal Data Science team. However, it faced a significant challenge: many models in production were affected by an outdated and erroneous feature store, which had a negative impact on their performance and compliance with SLAs. Inconsistencies in the delivery of predictions, often late or incorrect, exposed the urgent need for specialized intervention. The MLops team, in charge of maintaining and updating the feature store (known internally as Data Engine), was saturated and looking for solutions.
WhiteBox responded to this need by integrating an expert Machine Learning engineer into the Vodafone MLOps team. As another member of the team, this professional brought with them enormous technical experience, leading innovation in the generation of new features and the modernization of the feature store modules. This renewal, which was based on advanced Big Data technologies through Google Cloud Platform (Dataproc), marks a turning point in managing the lifecycle of models at Vodafone.
The contribution of our Machine Learning Engineer resulted in tangible changes to Vodafone, allowing it to overcome obstacles and efficiently update its models. The quality and reliability of the feature store experienced a substantial improvement, freeing up the team of data scientists to focus on refining the models. As a result, there was a notable increase in the productivity of the department, with models that operated more quickly and with greater precision, translating into benefits for the business.
- Feature store with more than 5,000 features.
