Amazon Web Services
Senior UX Designer · January 2025
Project Entity Tables
Redesigning discovery and access to project assets in Amazon SageMaker Unified Studio, creating a scalable inventory system so data builders can instantly find everything they've created, from one place.
Product
Amazon SageMaker Unified Studio
Team
UXD, PM, Engineering, Data
Delivered
January 2025 · Engineering handoff complete
← All work
01 · Context
A unified platform for the full data stack
Amazon SageMaker Unified Studio (SMUS) is a unified development environment where data workers can build, deploy, process, and analyze data-driven products, bringing together AWS analytics, ML, and generative AI tools under one roof.
The platform organizes work through a clear hierarchy: Domains contain Projects, and projects contain Tools, the specific AWS capabilities a team has enabled. Each tool lets builders create entities: apps, models, workflows, pipelines, and more.
Level 1
Domain
Level 2
Project
Level 3 · Tools
Query Editor
Query Book
Workflows
SageMaker Studio
Bedrock Studio
Quicksite
How it works
SMUS is where data teams go to do their actual work. It consolidates familiar AWS tools (Amazon EMR, Glue, Athena, Redshift, Bedrock, and SageMaker AI) into a single governed environment, so teams don't have to jump between disconnected consoles to get things done.
In practice, that means a data engineer can query a data lake, a machine learning engineer can train and deploy a model, and a GenAI builder can create a Bedrock-powered application, all from the same place, within the same project, with access to the same data.
Work is organized into projects: shared spaces where team members collaborate, create artifacts, and share analytics and AI assets across the organization. Each project is scoped to a set of tools (defined by its project profile), and everything a team builds (notebooks, pipelines, prompts, models, applications) lives inside that project.
The core use cases span four areas: SQL analytics for querying and exploring data, data processing for building ETL pipelines and workflows, model development for training and deploying ML models, and generative AI for building applications powered by Amazon Bedrock. These workflows increasingly overlap, which is exactly why a unified environment exists in the first place.
SMUS Personas
Three primary builder types use the platform, each with distinct goals and key actions.
Data Explorer
“I discover and explore data assets in my organization.”
Discovers ML assets, requests access, explores shared data. Key CTAs: Discover data · ML model hub · GenAI playground.
Data Product Builder
“I build scripts, workflows, and pipelines for data assets.”
Connects to project data, cleans and preps, builds pipelines and workflows. Key CTAs: Browse data · Create data product.
GenAI & ML Builder
“I build and fine-tune ML models and GenAI apps.”
Trains ML models, builds GenAI apps from ML models. Key CTAs: ML model hub · GenAI playground · Create GenAI app.
Three primary builder types use the platform, each with distinct goals, workflows, and key actions.
🔍 Click image to enlarge
Data admins create domains in the AWS console. Data explorers and builders then launch these domains in an off-console portal to create projects for data analysis, AI model development, and application building.
02 · Problem
Where's my stuff?
A problem occurred when builders who had previously created entities in a project could not find those entities upon landing on the project overview page. The project creation workflow and project overview APIs prevented entities and other tool assets from surfacing at the top level, making them difficult to find.
Core Friction
Invisible entities
Project entities created inside tools (Bedrock apps, workflows, ML pipelines) didn't surface on the project overview page. Users couldn't find what they built.
User Impact
Split between Git and local
Local artifacts like Bedrock assets weren't visible because they weren't saved to the Git repo. Users needed visibility of both Git and local artifacts in one view.
Scale Risk
No inventory as tools multiply
With each new tool category added to SMUS, the problem compounds. Without a scalable entity display system, the overview page would become increasingly useless.
“Where're our GenAI app components? I created assets in Bedrock, but I can’t find them anywhere.”
Local artifacts (like Bedrock assets) weren't visible on the project overview page since they weren't saved to the Git repo. Users needed visibility of both Git and local artifacts.
The project overview page before the redesign. Only a Git repo file list was visible. No project entities, no quick access to tools, and no way to discover assets created across different tools.
We simply needed a more discoverable and scalable inventory display of project entities to solve the immediate need to find and open files or assets from the overview page.
05 · Design Proposals
Entity tables conditioned on project configuration
The core design decision: only surface entity tables for tool categories that were enabled at project creation. This keeps the overview page clean and relevant, never showing empty categories that don't apply to the project.
A single table per category shows 6 entities per page with pagination, a category filter dropdown, and a "Create Entity" button always in view.
The very first element: a Get Started banner with actionable shortcut buttons to take users directly into the tools created in the project: Build chat agent app, Build prompt flow, Manage prompts, and Model evaluation.
The full-project state: all five entity categories visible with their respective tables, pagination, filter, and Create Entity actions.
A project created with only Amazon Bedrock: only GenAI application development entities are displayed: Chats, Prompts, Functions, Guardrails, and Knowledge bases.
🔍 Click image to enlarge
Before: users returned to the project overview and couldn't find their entities ("Where's my file?"). After: entities are visible directly on the overview page, allowing users to resume work instantly ("Here's my file!").
