Exowatt beats frontier models on accuracy with SageMaker serverless model customization
In this post, we show how Exowatt used Amazon SageMaker AI serverless model customization to build a repeatable AI pipeline that produces task-specific fine-tuned models for two deployment targets, edge and cloud, achieving higher accuracy than frontier models at up to 49x lower cost.
Exowatt, a company specializing in modular solar systems for high-energy applications like AI data centers, has achieved higher accuracy in its AI models using Amazon SageMaker AI serverless model customization. Their flagship product, Exowatt P3, captures solar energy, stores it as heat, and dispatches electricity without relying on the grid. This innovation allows Exowatt to power data centers directly, eliminating the need for multi-year grid interconnection waiting periods.
To create highly accurate, task-specific models for edge and cloud deployment, Exowatt utilized Amazon SageMaker AI serverless model customization. This methodology generates fine-tuned models that outperform frontier models while reducing costs by up to 49 times. The two primary models tested were Qwen3-1.7B for edge alarm triage and Qwen3-14B for cloud energy-market analysis.
By adopting a serverless model customization approach, Exowatt built a repeatable AI pipeline that starts with raw operational data and culminates in a certified training set, fine-tuned model, and evaluated deployment artifact. This pipeline enables Exowatt to transition from weeks to days in the model development process. Alarm triage and energy-market analysis are just two of many narrow, high-value tasks where this approach can be applied.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
