Pioneering GenOps: Vodafone’s Cutting-Edge Platform on Google Cloud

The collaboration has been performance transformational, with:

• Two-week Machine Learning model time to deployment, reduced from three months
• 600+ Machine Learning models enabled into production
• 70%+ average reduced manual effort through code reuse

 

How it works

Vodafone’s AI Booster platform is a cutting-edge Machine Learning Ops platform based on Google Cloud architecture, with built-in automation, scalability, and security. Developed in collaboration with Datatonic and Google, it uses Vertex AI to streamline the deployment of machine learning models. AI Booster serves over 3,000 data scientists across Vodafone’s markets and multiple lines of business, including B2B, Commercial, Finance, Customer Care, and Supply Chain Management. 

With Generative AI offering new and innovative ways to transform businesses, Vodafone wanted to expand its AI Booster platform by adding Generative AI capabilities. This will enable Vodafone to rapidly develop and deploy new GenAI use cases while being fully compliant with responsible AI principles.

“Fundamentally, safety is the most important focus for how we work and for all our customer interactions. This means solid moderation and checks to protect against inappropriate and biased content, and robust anti-jail-break and anti-prompt injection controls. Vodafone is supporting the full lifecycle not just the applicative software, we take accountability for the knowledge base and the LLM, having introduced automatic benchmarking on a set of golden reference prompts,” – Pierguido Vincenzo Carlo Caironi, Vodafone Head of Data Portfolio

 

The approach

Vodafone, Datatonic, and Google Cloud have evolved AI Booster into a best-in-class GenOps platform, enabling a rapid path to production for GenAI use cases. Three key requirements act as underlying principles for the technical build: 

1. Compliance with the EU AI Act
2. Automation capabilities for GenAI with CI/CD
3. Multi-cloud tenancy support so models from different clouds can be used, enabling Vodafone teams to choose the best models for their use cases

 

Many platform capabilities wrap around the deployed use cases:

• Hosting of LLMs across clouds provides freedom of choice among the competitive and evolving LLM landscape to choose the best performing model for a use case while ensuring highest security and governance standards.
• Accelerating the time it takes to vectorize, retrieve, and serve insights from data is enabled with a templated RAG solution.
• Compliancy with Vodafone’s high security standards, grounding responses in trusted enterprise data, and ensuring Responsible AI principles by masking sensitive information, avoiding common malicious attacks such as prompt injection, and moderating content.
• Closely monitoring and visualising the performance of GenAI use cases by applying MLOps best practices such as continuous model monitoring and evaluation combined with LLM-specific metric computations.

 

The impact

With the new enhancements to the platform, Vodafone have already deployed high impact use cases such as:

1. Ask HR Chatbot to provide specific detailed answers to employee queries on HR guidelines and procedures. 
2. Proposal Assistant to win additional business covering all the bids raised by Vodafone’s business customers. 
3. Marketing Content Creation to generate content faster and in-house. 

AI Booster is delivering significant value to Vodafone, such as reducing Machine Learning model time to deployment to two weeks from three months, as well as optimising cost savings, driving improved efficiency, and executing on enhanced security. 

“Datatonic is proud to work alongside Vodafone and Google Cloud on this groundbreaking GenOps project. Our collaboration has resulted in a truly innovative platform that empowers Vodafone to harness the power of generative AI responsibly and at scale. We are excited to continue supporting Vodafone’s AI journey and look forward to helping other leading businesses unlock the incredible potential of GenAI use cases.” – Louis Decuypere, CEO of Datatonic