Title

AWS re:Invent 2023 - Democratize ML with no code/low code using Amazon SageMaker Canvas (AIM217)

Summary

• Rajneesh Singh, GM of SageMaker low code/no code team, along with co-speakers David De Gallitelli and Ramdev Udali, discussed democratizing ML with SageMaker Canvas.
• The session covered new features in SageMaker Canvas, including demos and a case study from Thomson Reuters.
• SageMaker Canvas is a no-code workspace for building, deploying ML, and generative AI models, offering pre-trained models and custom model building capabilities.
• It supports collaboration with ML experts through SageMaker Studio and provides insights into data with built-in visualizations and transformations.
• New features include visual data preparation, natural language data prep, and fine-tuning of foundation models.
• The session included a demo of fine-tuning a generative AI model and building a custom model with data preparation and model training.
• Ramdev Udali shared how Thomson Reuters uses SageMaker Canvas for hackathons and empowering citizen developers.
• Resources for learning and further exploration of SageMaker Canvas were provided, including a Coursera course and hands-on labs.

Insights

• The demand for ML in organizations is growing, but there is a shortage of ML experts, leading to a need for no-code/low-code solutions like SageMaker Canvas.
• SageMaker Canvas is designed to empower non-technical domain experts to participate in ML projects, fostering collaboration and speeding up the ML initiative rollout.
• The integration of pre-trained models from AWS AI services, such as Amazon Bedrock, Textract, Comprehend, and Recognition, simplifies the process of applying ML to various business problems.
• The ability to fine-tune foundation models within Canvas with labeled data and evaluate them for responsible AI metrics like toxicity and bias is a significant advancement.
• The case study from Thomson Reuters illustrates the practical application of SageMaker Canvas in an enterprise setting, highlighting its effectiveness in enabling non-ML experts to build and deploy ML models successfully.
• The session's emphasis on new features and the live demo underscore AWS's commitment to making ML more accessible and reducing the barrier to entry for businesses looking to implement ML solutions.
• The provision of educational resources and additional sessions for deeper dives into SageMaker Canvas features indicates a strong support ecosystem for users of all levels.