SageMaker Canvas Ready-To-Use AI Models

Generate insights with pre-built models without writing a single line of code

the challenge

Organizations across manufacturing, healthcare, finance, and other industries face significant barriers in leveraging AI/ML to improve efficiency and derive insights from their data. The primary challenge is that line-of-business users must rely on scarce AI/ML practitioners and developers to build solutions, even for simple proof-of-concepts, creating bottlenecks that can delay projects by months. This dependency is particularly problematic when dealing with business data requiring customization, as non-technical users lack tools to build their own custom models.

Additionally, while hundreds of pre-trained AI models and managed services exist, accessing and implementing these solutions still requires technical expertise that most business users don't possess. As a result, organizations struggle to realize AI value quickly, improve team productivity, and enhance operational efficiency due to these technical barriers and resource constraints.

ROLE

Lead UX designer

DURATION

3 months (January 2023 - March 2023)

KEY RESPONSIBILITIES

  • Performed competitive analysis and shared insights with stakeholders to inform product requirements

  • Participated in customer sessions to understand pain points, iterated on designs based on feedback, and uncovered new opportunities to innovate

  • Partnered with PM and Engineering to prioritize MLP requirements and inform future roadmap

TEAM

  • UX: 1

  • PM: 1

  • Engineering: 6 (1 SDM, 3 FEs, 2 BEs)

DESIGN PROCESS

research

During the research phase, I partnered with the PM to understand existing customer pain points and define personas across key use cases. These users can be business analysts, financial analysts, customer support managers, or insurance claim officers who are tasked with classifying, analyzing, or extracting information from documents, text, and images manually. With a clear picture of user and customer needs, we collaborated to establish product requirements and articulate the jobs to be done.

We generate thousands of clinical documents annually, each of which must be assessed by clinical experts for compliance through manual processes.

Director of IT

Clinical genetic testing company

Our data science team was tasked with building these models but was reliant on line of business users as they are the data experts, and the process took months.

Head of Digital Innovation

Global energy technology company

Our data science team was tasked with building these models but was reliant on line of business users as they are the data experts, and the process took months.

Head of Digital Innovation

Global energy technology company

We need to classify free-form text description of surgical instruments used by physicians but do not have the expertise to build our own text classifier.

Head of Analytics

Healthcare technology company

We need to classify free-form text description of surgical instruments used by physicians but do not have the expertise to build our own text classifier.

Head of Analytics

Healthcare technology company

I evaluated existing AWS services offering classification and insight analysis from documents, text, and images, identified key usability gaps, and consolidated the findings into a centralized document shared with stakeholders.

concept iteration

During the iteration phase, I developed multiple concepts, collaborated with stakeholders, and aligned on the direction that best addressed customer needs.

launched features

Landing page

Users can discover and browse through all pre-built AI solutions. They can interactively evaluate how each solution works on a single image, document, or line of text. Users can start using the solution by uploading data to be analyzed. The solution will apply AI to the user uploaded data and generate predictions for the user to review.

Sentiment analysis

Sentiment analysis allows users to detect the sentiment in lines of text. The predicted results can be positive, negative, neutral, or mixed, giving users a reliable read on sentiment without needing technical expertise or custom ML models.

Entities extraction

Entities extraction allows users to extract entities such as people, places, and commercial items, or units such as dates and quantities, from text. The predicted results are the entities extracted from the text, allowing users to turn unstructured content into usable data without building custom NLP models.

Language detection

Language detection allows users to determine the dominant language in text. The predicted results are the languages detected in the text, enabling instant multilingual routing without specialized expertise.

Personal information detection

Personal information detection allows users to detect personal information that could be used to identify an individual such as addresses, bank account numbers, and phone numbers, from text. The predicted results are the entities of personal information extracted from the text, helping users protect sensitive data without needing custom detection tools.

Object detection in images

Object detection in images allows users to detect objects, concepts, scenes, and actions in an image. The predicted results are the different objects detected from the image, giving users quick visual insights without needing custom vision models.

Text detection in images

Text detection in images allows users to detect text in an image. The predicted results are the different text labels detected from the image, turning visual content into usable text without requiring technical expertise.

Expense analysis in invoices

Expense analysis in invoices allows users to extract information from invoices such as date, number, item prices, total amount, and payment terms. The predicted results are the different summary and line item fields detected from the invoice, enabling fast, automated invoice processing without custom AI models.

Expense analysis in receipts

Expense analysis in receipts allows users to extract information from receipts such as date, number, item prices, total amount, and payment terms. The predicted results are the different summary and line item fields detected from the receipt, enabling quick, automated receipt processing without custom AI models.

Identity document analysis

Identity document analysis allows users to extract information from passports, driver licenses, and other identity documentation issued by the US government. The predicted results are the different fields and values detected from the identification document, enabling fast verification without the need for custom AI models.

Document analysis

Document analysis allows users to analyze documents and forms for relationships among detected text. The predicted results are the raw text, forms, tables, and signatures detected from the document, giving users structured data without needing custom document‑processing models.

Document queries

Document queries allows users to extract information from structured documents such as paystubs, bank statements, W‑2s, and mortgage application forms by asking questions using natural language. The predicted results are the values detected from the document, giving users quick answers without needing custom parsing or ML expertise.

RESULTS AND IMPACT

GA launch

GA launch

In March 2023

In March 2023

Immediate impact

Pressure-tested patterns that scale across all solution types within a week

Immediate impact

Pressure-tested patterns that scale across all solution types within a week

Scaled to 50+ models

Defined a modular pattern that scaled to 50+ ready-to-use models, eliminating the need for users to learn new patterns for each solution

Scaled to 50+ models

Defined a modular pattern that scaled to 50+ ready-to-use models, eliminating the need for users to learn new patterns for each solution

SageMaker Canvas allows our clinical experts to now extract insights and data from thousands of documents without even thinking about building a model.

Director of IT

Clinical genetic testing company

Image classification in Canvas has allowed us to scale the usage of computer vision models and solve dozens of use cases in a fraction of the time.

Head of Digital Innovation

Global energy technology company

Image classification in Canvas has allowed us to scale the usage of computer vision models and solve dozens of use cases in a fraction of the time.

Head of Digital Innovation

Global energy technology company

The no-code platform now enables us to run predictions on more than 10K rows of text in matter of minutes without the need of any technical expertise.

Head of Analytics

Healthcare technology company

The no-code platform now enables us to run predictions on more than 10K rows of text in matter of minutes without the need of any technical expertise.

Head of Analytics

Healthcare technology company

DEMO

other projects

© 2026 Wayne Chou • UX + Visual Design. All Rights Reserved.

© 2026 Wayne Chou • UX + Visual Design. All Rights Reserved.

© 2026 Wayne Chou • UX + Visual Design. All Rights Reserved.

Create a free website with Framer, the website builder loved by startups, designers and agencies.