Unlocking AI Transformation: Hands-On Learning with Atos and AWS AI League

Empowering Workforce Upskilling through Gamified Experiences

Bridge the Gap: From Theory to Practical AI Application

Accelerating AI Education: The Role of Experiential Learning

The AWS AI League: Revolutionizing AI Skill Development

Fine-Tuning: A Strategic Advantage for Business Use Cases

A Deep Dive into the AWS AI League Framework

Real-World Impact: The Intelligent Insurance Underwriter

Simplifying Fine-Tuning with Amazon SageMaker

Elevating Engagement through Gamification

Best Practices for Successful Fine-Tuning Strategies

Overcoming Challenges: Understanding Model Performance

Achieving AI Fluency: Atos’s Journey Towards 2026 Goals

Conclusion: The Future of AI Learning and Application

Meet the Authors: Insights from Industry Experts

Accelerating AI Education at Scale: Lessons from Atos’ AWS AI League Experience

This post is co-written with Mark Ross from Atos.

Organizations pursuing AI transformation often encounter a significant challenge: how to upskill their workforce effectively and at scale. Traditional training methods—whether online courses, certification programs, or classroom instruction—serve as necessary foundations but frequently fall short. They often fail to engage teams sufficiently, leading to a disconnect between theoretical knowledge and real-world application. This means certification does not always translate to confidence or competence in addressing business challenges with AI.

Recognizing this gap, Atos, in partnership with AWS, has focused on hands-on learning as the key to effective AI enablement. By integrating structured e-learning with experiential training, they aim to transform theoretical knowledge into actionable skills. With over 5,800 AWS Certifications and 11 Golden Jackets under their belt, Atos is committed to achieving a 100% AI-fluent workforce by 2026. This post explores how Atos utilized the AWS AI League to scale engagement, accelerate practical skills, and empower engineers.

AI Enablement through the AWS AI League

E-learning courses form the backbone of AI education, but many organizations still struggle to bridge the gap between knowledge and practical application. The AWS AI League addresses this challenge by emphasizing hands-on experimentation alongside structured competition. Participants engage with generative AI tools in real-world scenarios, fostering collaboration and measurable outcomes.

Through this innovative program, Atos enables participants to fine-tune large language models (LLMs) via platforms like Amazon SageMaker and SageMaker JumpStart. This hands-on experience prepares teams for successful enterprise AI adoption, helping them move beyond passive learning.

Why Fine-Tuning Matters for Business Use Cases

Fine-tuning is a crucial aspect of adapting pre-trained models to specific domains. This approach minimizes training time and costs, while ensuring models encapsulate domain-specific knowledge. For Atos, examples include fine-tuning models for insurance underwriting, where intricate risk assessment and policy recommendations demand more than generic language capabilities.

With the AWS AI League’s support, professionals across various specialties—including solution architects and business analysts—now have the ability to customize and deploy powerful AI models without needing deep machine learning expertise.

How the AWS AI League Works

The AWS AI League follows a three-stage structure that emphasizes hands-on learning and engagement:

1. Immersive Workshops: Participants gain foundational knowledge and engage with SageMaker JumpStart to explore fine-tuning.

2. Model Development Phase: Teams experiment with various fine-tuning strategies, continuously improving their models while competing on a leaderboard that reflects real-time performance.

3. Interactive Finale: Top teams demonstrate their models through real-time challenges, evaluated by technical judges and audience votes.

This structure emphasizes experimentation, visibility of progress, and model performance, reinforcing the goal of producing actionable, real-world solutions.

Atos’s Use Case – Intelligent Insurance Underwriter

To illustrate the AWS AI League’s impact, Atos developed the Intelligent Insurance Underwriter use case. Participants created a model capable of analyzing complex insurance scenarios, providing risk assessments and guidance on policy conditions. This approach goes beyond theoretical exercises—it’s a practical application of generative AI for real-world challenges.

Built on Amazon SageMaker and complementary AWS services, the solution combines a knowledge base with reasoning modules that draw from proprietary underwriting data. As a result, Atos developed an effective assistant that aids underwriting professionals, ensuring both efficiency and industry compliance.

Fine-Tuning with Amazon SageMaker Studio and JumpStart

Participants conducted model fine-tuning within Amazon SageMaker Studio, utilizing its integrated environment to create, adjust, and deploy generative AI models with ease. They selected from a catalog of pre-trained models, connected training datasets from Amazon S3, and managed the fine-tuning process via automated tools. This significantly reduces the complexity of operationalizing specialized AI models while maintaining performance efficiency.

Users iterated on hyperparameters such as learning rate and epochs, focusing on optimizing their models for performance. This systematic approach aided participants in identifying the best configurations for their use cases.

Gamification Ignites Participation

Atos observed remarkable engagement levels during the AWS AI League, with 409 participants competing and over 4,100 models created. The gamified format encouraged collaboration and healthy competition, where participants aimed to improve their scores on a dynamic leaderboard.

Communication was vital, as participants balanced support with self-discovery, enhancing their learning experience. The top contenders advanced to a live gameshow finale, showcasing enhanced model performance in an exhilarating environment.

Tips to Fine-Tune Your Way to Success

To achieve success in the competition, participants focused on two critical elements: generating a robust dataset and optimizing hyperparameters. Some even created customized tools inspired by AWS applications, enhancing the diversity and quality of their datasets.

Key strategies included:

• Utilizing Generative AI Tools: Enhance dataset diversity with alternative models and formats.
• Systematic Hyperparameter Tuning: Experiment with epochs, learning rates, and other parameters to optimize outcomes.
• Monitoring Loss and Perplexity: Keep track of evaluation metrics to avoid overfitting and ensure generalization capability.

Upskilling Ambitions Achieved

The AWS AI League significantly advanced Atos’s ambition of empowering its workforce with generative AI capabilities. Participants learned how targeted fine-tuning with a smaller model could outperform more extensive models, proving that domain-specific knowledge is invaluable. As organizations look to implement AI across various industries, cost-effective, fine-tuned models will be crucial.

Conclusion

As we navigate the evolving landscape of generative AI, the ability to effectively build and deploy specialized models becomes increasingly vital. The AWS AI League serves as a roadmap for organizations like Atos, enabling the development of AI-driven solutions that address real-world challenges. This gamified learning experience not only accelerates innovation but also drives tangible business outcomes.

If you’re interested in enhancing your AI capabilities through experiential learning, consider hosting your own AWS AI League event.

For more insights on implementing AI solutions, check out the AWS Artificial Intelligence blog for stories about industry pioneers leveraging generative AI.

About the Authors

Nick McCarthy is a Senior Generative AI Specialist Solutions Architect with a focus on helping customers customize their GenAI models with AWS.

Mark Ross is the Chief Architect for AWS within Atos, leveraging over two decades of technology experience to assist clients in navigating the AWS landscape.

This post highlights a forward-thinking approach to AI upskilling, showcasing how organizations can harness hands-on, gamified learning to transform their workforce and drive innovation.