AWS re:Invent 2025 - Democratizing Whirlpool's Virtual Product Development with AWS (IND331)

Democratizing Whirlpool's Virtual Product Development with AWS

Introduction to Whirlpool

• Whirlpool is a leading home appliance company with $17 billion in annual sales and 44,000 employees globally
• They have 40 manufacturing and product development/research centers worldwide
• Their iconic brands include Whirlpool, KitchenAid, Maytag, Amana, and others

Challenges in Product Engineering

• Products are becoming increasingly connected and smart, requiring integration of mechanical, electrical, and software design
• This complexity leads to longer product development timelines
• There is constant pressure to reduce costs and optimize product quality

How AWS Enables Transformation

• AWS offers over 200 services that can help manufacturers modernize their product engineering processes
• Key capabilities include:
  • Scalable infrastructure to handle fluctuating simulation and testing needs
  • Access to HPC, AI/ML, and other advanced capabilities to leverage product data
  • Flexibility to optimize costs and infrastructure for engineering workloads
  • Ability to unlock data silos and enable more integrated workflows

Whirlpool's Virtual Product Development Approach

Purpose, Scope, and Implementation

• Translating system-level requirements into an optimal product architecture is challenging
• Traditional methods are fragmented and often lead to suboptimal solutions
• Whirlpool uses a multi-objective optimization framework to balance factors like energy, capacity, robustness, noise, and cost

Simulation Model Validation

• Whirlpool has a structured process to validate simulation models at different levels of fidelity:
  • Developmental models to assess simulability
  • Directional models to understand sensitivities
  • Quantifiable models that bridge to validated models
  • Validated models that can replace physical testing
• Validation ensures high accuracy and stakeholder alignment on using simulation to drive decisions

Key Performance Metrics

• Internal capacity: Matching CAD and physical measurements
• Coefficient of performance: Balancing heat gain and energy consumption
• Robustness: Minimizing external sweating, frost formation, and internal condensation
• Cost: Optimizing insulation and cooling component costs

Design Space Exploration

• Whirlpool's product has over 34 design factors with 2-5 levels each, leading to over 5.5 trillion possible combinations
• They use Design of Experiments (DOE) techniques like orthogonal arrays to efficiently sample the design space
• This allows them to fit accurate machine learning models with just ~1,000 runs, rather than exhaustively exploring the full design space

Optimization and Decision-Making

• Whirlpool uses multi-objective genetic algorithms to optimize the product design
• Key steps include:
  1. Generating a constrained population of feasible design points
  2. Predicting performance using the ML models
  3. Evolutionary selection to explore the Pareto-optimal solutions
  4. Selecting the final design based on weighted objectives

AWS Architecture for Virtual Product Development

• Whirlpool has built a modular, serverless architecture on AWS to enable this virtual product development workflow:
  • API Gateway and Lambda functions orchestrate the end-to-end process
  • Custom Docker container in ECR handles model training, tuning, and optimization
  • SageMaker is used for scalable model training and hyperparameter tuning
  • S3 stores model artifacts and other data resources consumed by the front-end
• This architecture allows Whirlpool to efficiently manage the model lifecycle and make the capabilities available to their engineering teams through a user-friendly front-end application

Business Impact

• Whirlpool estimates up to 70% efficiency improvements in model training and tuning using this approach
• They've achieved at least 95% accuracy in defining the optimal product architecture
• This allows them to significantly accelerate decision-making in the early stages of product development
• The modular, serverless architecture provides flexibility and scalability to handle complex engineering workloads

Conclusion

Whirlpool's virtual product development approach, powered by AWS, enables them to:

• Efficiently explore a vast design space using advanced simulation and optimization techniques
• Validate simulation models to replace physical testing and accelerate the engineering process
• Leverage AWS services to build a scalable, flexible, and user-friendly platform for their engineers
• Achieve significant improvements in development efficiency, effectiveness, and product quality

This comprehensive, cloud-based solution helps Whirlpool define winning product architectures and stay competitive in the rapidly evolving home appliance market.