Denis Batalov

AI agents are increasingly making real-world decisions — approving loans, issuing refunds, offering discounts — yet most fairness testing still focuses on whether chatbots use offensive language. This gap creates real business and regulatory risk.

The rapid growth of generative AI brings promising innovation but raises new challenges around its safe and responsible development and use. While challenges like bias and explainability were common before generative AI, large language models bring new challenges like hallucination and toxicity.

Risk assessment is an essential part of developing AI solutions responsibly, especially with emerging industry standards and laws regarding AI risk, such as ISO 42001 and the EU AI Act.

Foundation models (FMs) can be used for natural language processing tasks such as summarization, text generation, classification, open-ended Q&A, and information extraction.

Amazon SageMaker brings together fully managed infrastructure, tools, and workflows so you have everything you need to prepare data and build, train, and deploy machine learning (ML) models. SageMaker offers many innovative features such as an IDE for ML and interfaces for business analysts to generate predictions without code.

AWS offers many choices for solving business problems through machine learning (ML), ranging from built-in algorithms, frameworks, and more in ML services down to models and algorithms available via AWS Marketplace. Amazon SageMaker supports 17 built-in ML algorithms, such as classification, regression, and recommendation.

Amazon SageMaker Ground Truth makes it easy to quickly label high-quality, accurate training datasets. In this workshop, we set up labeling jobs for text and images to help you understand how to make the most of Amazon SageMaker Ground Truth.

This presentation was recorded prior to re:Invent. In this session, we showcase the power of machine learning in taking a large corpus of a foreign language text (we use the entire Wikipedia) and automatically learning word embeddings for that language.