Training Azerbaijani language models on Amazon SageMaker AI
Azercell Telecom LLC, Azerbaijan's leading telecom provider, collaborated with the AWS Generative AI Innovation Center over six weeks to train an Azerbaijani large language model (LLM) on Amazon SageMaker AI. The project aimed to adapt foundation models to Azerbaijani, a morphologically rich language with scarce training data and no existing blueprint for efficient LLM training. The resulting production-ready framework supports telecom use cases, including a customer-facing chatbot. This work demonstrates how SageMaker AI can be leveraged for low-resource language modeling, providing a template for similar initiatives. The framework is now available for Azercell's deployment, potentially improving customer service and operational efficiency.
Enables developers to train LLMs for low-resource languages using SageMaker AI.