Publishing tool for datasets and models

Publishing tool for datasets and models

UX/UICompetitor analysisInterview

Task

  • Develop a publishing tool for datasets and models for internal and external ML and Data Science specialists
  • Account for the needs of both audiences — external and internal
  • Ensure flexible filtering
  • Support both public and private placements
  • Lay the groundwork for future monetization

What was done

  • Conducted a competitive analysis: studied Hugging Face, Kaggle, data.world, and Amazon Data Marketplace — their concepts, target audiences, key features, monetization methods, and UI patterns
  • Presented the analysis results to the team
  • Ran a series of interviews with ML and DS specialists to find out which services they use, what criteria they use to search for models and datasets, which filters matter most, and what data they consider critical on the pages
  • Designed the mockups based on these insights using components from the Malachite design system (SberBusiness): dataset and model description pages, catalogs with filters and sorting, and interfaces for creating and editing datasets

Result

  • Designed scalable catalog interfaces for datasets and models based on real ML and Data Science specialists' workflows
  • Expected early-stage metrics: 500–1,000 active users per month, 50–150 published datasets, 30–80 published models
Competitors page structure analysis
Competitors page structure analysis
Datasets catalog and dataset page
Datasets catalog and dataset page
Models catalog and model page
Models catalog and model page
Screens for creating a model or dataset
Screens for creating a model or dataset
Drawn layouts
Drawn layouts
Next caseModerated user interview