01
Robovision
Centralizing scattered data workflows to reduce confusion and unlock hidden platform capabilities
Centralizing scattered data workflows to reduce confusion and unlock hidden platform capabilities
Client
Client
Robovision
Robovision
Type
Type
Feature design
Feature design
Role
Role
Lead Product Designer
Lead Product Designer
Year
Year
2025
2025
The challenge
The challenge
The Label Center suffered from feature creep over several product iterations. Each new capability (2D/3D labeling, special imports, AI labeling, test data ingestion) was added to solve immediate needs, but gradually transformed the Label Center from a focused workspace into an overloaded system that users found increasingly difficult to navigate.
60%
Reduce time spent searching for transformed data across the platform
80%
Decrease support requests about "missing" features that already exist
100%
Centralize all data outputs in one discoverable location
Research & findings
Research & findings
Data scattered Across Platform
Data scattered Across Platform
Users transformed data in one area but couldn't find it later when working in different sections. This led to duplicate work and frustration with the platform's perceived limitations.
Users transformed data in one area but couldn't find it later when working in different sections. This led to duplicate work and frustration with the platform's perceived limitations.
Hidden Feature Discovery Problem
Hidden Feature Discovery Problem
Users repeatedly requested features that already existed. They simply couldn't discover them because data outputs weren't visible or linked across workflows.
Users repeatedly requested features that already existed. They simply couldn't discover them because data outputs weren't visible or linked across workflows.
Mental Model Mismatch
Mental Model Mismatch
The platform structure didn't match how data scientists think about their work. They expect a centralized data workspace, not isolated transformation silos.
The platform structure didn't match how data scientists think about their work. They expect a centralized data workspace, not isolated transformation silos.
Power User Resistance Risk
Power User Resistance Risk
Long-time users had adapted workarounds. Any restructuring needed to preserve their workflows while improving discoverability for new and intermediate users.
Long-time users had adapted workarounds. Any restructuring needed to preserve their workflows while improving discoverability for new and intermediate users.
Design Principles
Design Principles
The Solution
The Solution
A centralized Data Management Center that brings all transformed data outputs into one discoverable workspace.
All transformations, regardless of where they were created now appear in a dedicated hub, allowing users to see their complete data landscape at a glance. Visual indicators reveal data lineage and relationships between transformations that were previously hidden. Users can still access data from their original work areas through clear links to the central hub, preserving power user workflows while adding visibility. Enhanced search and filtering make it easy to find specific outputs, with features surfaced contextually based on the user's current task.
A centralized Data Management Center that brings all transformed data outputs into one discoverable workspace.
All transformations, regardless of where they were created now appear in a dedicated hub, allowing users to see their complete data landscape at a glance. Visual indicators reveal data lineage and relationships between transformations that were previously hidden. Users can still access data from their original work areas through clear links to the central hub, preserving power user workflows while adding visibility. Enhanced search and filtering make it easy to find specific outputs, with features surfaced contextually based on the user's current task.
Information Architecture
Information Architecture
I focused heavily on information architecture to untangle the system and separate concerns clearly. All transformed data now lives in a centralized Data Center, searchable, well-categorized, and enriched with metadata. The transformation tools stay where users expect them but automatically link their outputs back to this hub. I also introduced a new top-level entry point so the Data Center is always one click away. Finally, I designed an intelligent search that lets users filter data by type, date, project, or transformation method, making discovery far easier.
I focused heavily on information architecture to untangle the system and separate concerns clearly. All transformed data now lives in a centralized Data Center, searchable, well-categorized, and enriched with metadata. The transformation tools stay where users expect them but automatically link their outputs back to this hub. I also introduced a new top-level entry point so the Data Center is always one click away. Finally, I designed an intelligent search that lets users filter data by type, date, project, or transformation method, making discovery far easier.
Key Learnings
Key Learnings
Small UX issues often reveal deeper gaps. What looked minor was actually blocking users from grasping the platform’s full value. I learned to improve existing workflows without disrupting the patterns power users depend on. In complex systems, information architecture is the design; clear mental models matter more than visuals. Early collaboration with my PM, UX lead, and engineers surfaced constraints that led to stronger, more realistic solutions. And I saw the importance of designing for all expertise levels.Simple by default, with depth available when needed.
Small UX issues often reveal deeper gaps. What looked minor was actually blocking users from grasping the platform’s full value. I learned to improve existing workflows without disrupting the patterns power users depend on. In complex systems, information architecture is the design; clear mental models matter more than visuals. Early collaboration with my PM, UX lead, and engineers surfaced constraints that led to stronger, more realistic solutions. And I saw the importance of designing for all expertise levels.Simple by default, with depth available when needed.
