Upskilling a global workforce in generative artificial intelligence requires more than just passive video lectures; it demands active, hands-on engagement. Atos recently tackled this challenge by collaborating with Amazon Web Services (AWS) to launch the AWS AI League, a competitive learning initiative designed to bridge the gap between theoretical knowledge and practical application. This program illustrates a growing trend in corporate education where gamification is used to drive high engagement levels among technical staff.

The program centered on an "Intelligent Insurance Underwriter" use case, where employees fine-tuned Large Language Models (LLMs) to handle complex, domain-specific tasks such as risk assessment and policy recommendations. Fine-tuning is a process where a pre-existing AI model is further trained on a specialized dataset to improve its accuracy in a particular niche. By utilizing Amazon SageMaker, a platform that provides the infrastructure for building and deploying AI, participants focused on data quality and hyperparameter tuning—the "knobs" that control how a model learns—rather than managing complex backend hardware.

The results were significant, with over 4,100 models generated by 400 participants during a two-week virtual sprint. To maintain high standards, the competition used "LLM-as-a-Judge," an automated system where a much larger model evaluates the responses of smaller ones. This approach mirrors real-world enterprise needs, where companies seek to deploy smaller, cost-effective models that still perform at expert levels. The initiative demonstrates that friendly competition can significantly accelerate AI literacy across diverse corporate roles while producing production-ready solutions.

Upskilling a global workforce in generative artificial intelligence requires more than just passive video lectures; it demands active, hands-on engagement. Atos recently tackled this challenge by collaborating with Amazon Web Services (AWS) to launch the AWS AI League, a competitive learning initiative designed to bridge the gap between theoretical knowledge and practical application. This program illustrates a growing trend in corporate education where gamification is used to drive high engagement levels among technical staff.

The program centered on an "Intelligent Insurance Underwriter" use case, where employees fine-tuned Large Language Models (LLMs) to handle complex, domain-specific tasks such as risk assessment and policy recommendations. Fine-tuning is a process where a pre-existing AI model is further trained on a specialized dataset to improve its accuracy in a particular niche. By utilizing Amazon SageMaker, a platform that provides the infrastructure for building and deploying AI, participants focused on data quality and hyperparameter tuning—the "knobs" that control how a model learns—rather than managing complex backend hardware.

The results were significant, with over 4,100 models generated by 400 participants during a two-week virtual sprint. To maintain high standards, the competition used "LLM-as-a-Judge," an automated system where a much larger model evaluates the responses of smaller ones. This approach mirrors real-world enterprise needs, where companies seek to deploy smaller, cost-effective models that still perform at expert levels. The initiative demonstrates that friendly competition can significantly accelerate AI literacy across diverse corporate roles while producing production-ready solutions.