AWS AI League: Atos fine-tunes approach to AI education

This article describes how Atos partnered with AWS to deliver the AWS AI League, a gamified learning program that accelerated AI education through hands-on fine-tuning of large language models.

• AWS AI League combines hands-on experimentation with structured competition for practical AI skills
• Atos used the program to upskill 400+ participants in fine-tuning LLMs using SageMaker and JumpStart
• Participants built an Intelligent Insurance Underwriter model for domain-specific underwriting guidance
• Fine-tuned 3B parameter models achieved 93%+ win rates against larger 90B parameter baseline models
• Gamification drove engagement: 409 participants created 4,100+ fine-tuned models over two weeks
• Key success factors: quality datasets, hyperparameter tuning, and avoiding overfitting on training data
• 85% of participants reported increased confidence implementing generative AI with customers post-event
• Fine-tuned models required less compute infrastructure, reducing inference costs at scale

The AWS AI League demonstrates how gamified, experiential learning accelerates practical AI skills and business impact compared to traditional training approaches.