Amazon SageMaker Canvas —

Meenakshisundaram Thandavarayan
8 min readOct 29, 2023

You mustn’t be afraid to dream up little bigger, darling…

Building generative AI applications has never been easier. Amazon SageMaker Canvas provides a no-code capability that democratizes generative AI across enterprises. You can now develop generative AI applications in minutes.

Amazon SageMaker Canvas is a no-code data science platform focused on making data science accessible to all. This post explores how the SageMaker Canvas team is working toward this ambitious vision of democratizing data science.

With Amazon SageMaker Canvas, users can leverage the breadth of AWS services for data, machine learning, artificial intelligence, generative AI, and business intelligence through a single interface. SageMaker Canvas abstracts away complexities, making these powerful technologies accessible to non-technical users. Whether it is building end to end ML lifecycle to comparing model performance or building RAG-based applications, SageMaker Canvas enables rapid development of generative AI solutions. By providing a no-code solution, Amazon SageMaker Canvas aims to spread the transformative potential of data science.

The body cannot live without the mind….

Amazon SageMaker is a Machine learning service designed to enable customers to build enterprise grade Machine Learning Platform. SageMaker offers 20+ sub services to address end to end machine learning capabilities. Let us break down SageMaker’s extensive capabilities into logical categories that showcase the full range of services for each ML life cycle stage

  • ML Life Cycle Services: Amazon SageMaker offers tools to cover each stage of the ML life cycle from data preparation to model development, evaluation, deployment, and inferencing. Example services include SageMaker Data Wrangler, SageMaker Estimators, SageMaker Processor, SageMaker Endpoint Configurations, and SageMaker Endpoints.
  • Responsible AI Services: Amazon SageMaker provides services to instill trust and monitor for issues throughout the ML life cycle. These services help identify bias in data and models, debug training processes, explain models post-training, and monitor models post-deployment. Example services include SageMaker Clarify, SageMaker Debugger, and SageMaker Model Monitor.
  • Artifact Management: Artifacts like feature data and model versions created during experiments and runs can be stored and managed in SageMaker Feature Store and SageMaker Model Registry. These services enable lineage tracking, version control, and approval workflows for artifacts.
  • Automation Services: SageMaker automation services like SageMaker Projects, Pipelines, and Experiments help scale ML efforts by automating, orchestrating, and industrializing ML workflows.
  • Governance Services: SageMaker governance features like Model Cards and MLOps Dashboards provide visibility, control, and governance across the ML life cycle from data to models to deployments.

May the Force be with you….

With AWS’ focus on “data science for all,” SageMaker Canvas represents a thoughtful engineering effort that packages SageMaker’s robust set of services to make them accessible to no code users. The complexities of data prep, model development, bias detection, explainability, and monitoring are abstracted away behind an intuitive interface. Users don’t need to be experts in SageMaker specifically or ML ops in general to develop, operationalize, and monitor models with SageMaker Canvas.

By making ML more consumable for non-technical users, SageMaker Canvas helps democratize data science across the enterprise. Users can tap into advanced ML capabilities that previously required specialized knowledge. Workflows that once spanned multiple services can now be accomplished through point-and-click interaction. The rich set of SageMaker services are now accessible to product managers, subject matter experts, and other personas that can benefit from applying ML to their work and insights.

Big things have small beginnings…

1. No Code Machine learning

Amazon SageMaker Canvas on its initial launch, started as a No Code Data science service enabling data and business analysts to quickly prepare their data sets, train a machine learning Model and evaluate model and do predictions. Features included but not limited to

  • Support for tabular data sets for Regression, Classification and time series forecasting
  • Exploratory Data Analysis — Bar plots, Scatter plots, Correlation Matrices
  • Data Validation and Preparation — traceable through recipe
  • Multiple model training modes — Quick build (2 to 10 minutes) for building a model for speed and Standard build (45 mins) for building a model for accuracy
  • Real time and Batch predictions

You can find additional information from AWS blogs such as this —

2. Data Connectors:

Machine learning models require substantial amounts of quality data to train effectively. AWS offers a range of data services and platforms to meet this need, from object stores like Amazon S3 to transactional databases in Amazon RDS to analytical databases such as Amazon Redshift.

To easily leverage these data sources for machine learning, SageMaker Canvas has built a connector framework that supports connecting to over 40 data sources. This enables both manual and automated ways to ingest data into SageMaker Canvas to build machine learning models.

Key data connection capabilities include:

  • Object Store Connectivity — Directly access data stored in Amazon S3 buckets
  • AWS Data Analytics Services Integration — Connect to Amazon Redshift data warehouses
  • AWS System of Record Databases — Link models to production databases like Amazon RDS
  • Partner Data Platform Support — Connect to partner data platforms like Snowflake and Databricks

The flexible data connectors allow data scientists to efficiently connect the latest AWS and partner data platforms to SageMaker Canvas for accelerated machine learning development.

3. Enhancing Accessibility with Single Sign-On

A key goal of SageMaker Canvas is to empower data analysts and business analysts to build machine learning models, even if they lack cloud expertise. To support this, SageMaker Canvas has implemented Single Sign-On (SSO) capability.

With SSO, users can directly access the SageMaker Canvas application without needing to log in to the AWS Management Console. This enhances accessibility, particularly for non-technical users who may find the AWS Console unfamiliar.

By exposing SageMaker Canvas via SSO, customers can get their data science teams quickly using the platform without cloud distractions. Whether connecting to AWS accounts or identity providers like Okta or Active Directory, SSO removes login friction that could hinder adoption.

The SSO integration exemplifies how SageMaker Canvas aims to open machine learning development to a broader range of users through improved accessibility and simplified workflows. By connecting SageMaker Canvas to existing identity systems, customers can accelerate their data science efforts while maintaining security and governance.

You can find detailed information here —

4. Seamless Model Development and Governance

To maintain oversight and enable continuous improvement, models built in SageMaker Canvas can be seamlessly shared with SageMaker Studio. Within Studio, models are

  • Versioned in the Model Registry — track model iterations and maintain lineage
  • Available for further evaluation — data scientists can thoroughly validate models before deployment
  • Open for feedback loops — data scientists can provide suggestions for improving the automated workflows in SageMaker Canvas

This integration between SageMaker Canvas and Studio enables a streamlined model development lifecycle. Automated workflows in Canvas rapidly generate candidate models. Human oversight in Studio then governs model quality through rigorous evaluation, versioning, and feedback loops back into Canvas.

5. Accelerating the Data-to-Insights Workflow

A key benefit of SageMaker Canvas is streamlining the end-to-end workflow from data to AI to business insights. This is enabled by seamless integration with business intelligence (BI) tools like Amazon QuickSight.

In a single click, batch predictions made in SageMaker Canvas can be exported to QuickSight data sets. This bridges the gap between model outputs and impactful visualizations that drive decisions.

With QuickSight integration, organizations can build true AI-powered BI capabilities, including:

  • Automated forecasts and projections based on machine learning models
  • Simulations and “what-if” analysis enhanced by AI
  • Interactive dashboards combining predictions with business data

The connectivity between SageMaker Canvas and QuickSight accelerates the data-to-insights cycle. Data analysts spend less time on manual handoffs between systems and more time delivering insights.

You can find more information here —

6. End-to-End Automation for Streamlined Machine Learning

SageMaker Canvas enables customers to automate their entire machine learning pipeline — from data to AI to business insights. With automations, any changes to the data source can trigger an automated workflow including:

  • Loading new data into SageMaker Canvas
  • Running predictions on the latest ML model
  • Pushing outputs into Amazon QuickSight for analysis

With SageMaker Canvas, customers can finally close the loop and create an integrated, automated workflow from raw data all the way to business impact. This accelerates time-to-value for machine learning investments.

You can find detailed information here —

7. Expanding Machine Learning Beyond Tabular Data

While tabular data is ubiquitous, Gartner studies show over 60% of enterprise data is unstructured in the form of text, images, videos and more. To unlock value, SageMaker Canvas integrates natively with AWS AI services for various data types:

  • Text Analysis — Amazon Comprehend for natural language processing and Amazon Textract for text extraction from documents
  • Image Recognition — Amazon Rekognition for identifying objects, people, text, inappropriate content and more in images and videos

These services provide pre-built models for common use cases. SageMaker Canvas combines these with the ability to build custom models tailored to unique needs.

By connecting multi-modal data sources into a unified workflow, SageMaker Canvas expands access to AI. The multi-modal capabilities make SageMaker Canvas suitable for diverse uniting use cases such as:

  • Sentiment analysis from customer surveys, reviews, call transcripts
  • Predictive search based on conversational language and images

8. Democratizing Generative AI

SageMaker Canvas provides easy access to cutting-edge generative AI models from Amazon, AWS partners, and open source hubs like Hugging Face. With no-code, users can:

  • Leverage Amazon and partner models as Bedrock serverless endpoints
  • Try out open source models hosted as SageMaker endpoints
  • Compare model performance side-by-side via chat interfaces
  • Build customized models with Amazon Bedrock using your own data

These features democratize generative AI, allowing all users to evaluate and benefit from the latest innovations.

For retrieval-augmented generation (RAG), SageMaker Canvas orchestrates connections between:

  • Retrieval systems like Amazon Kendra to aggregate relevant context
  • Generative models to produce human-like text and creative assets

RAG applications can be tailored to your use case by comparing multiple generation models. Connectors make it easy to index enterprise content in Kendra and enable permissions.

By simplifying access to diverse models, streamlining RAG building, and providing comparison guardrails, SageMaker Canvas unlocks generative AI’s potential while ensuring responsible use. Users of all skill levels can leverage these technologies to increase productivity and augment human creativity.

Cannot wait to see what new features are going to be release in 4 weeks at this year’s ReInvent. if you have not registered for the reinvent, you can get registered here —