Datarize
Frontend Engineer
Jan 2026 - Apr 2026
B2B SaaS platform for CRM marketing automation based on e-commerce customer data.
Increased widget-creation completion from 11% to 37% — built AI design generation from natural-language prompts
- Identified user abandonment caused by the effort of manually configuring design options during widget creation
- Designed an AI chat feature for generating and modifying widget designs from natural-language prompts, reducing manual configuration steps
- Improved the widget-design flow, increasing creation completion from 11% to 37% (approximately 3.4×)
Removed legacy dependencies across 30 components and 480 files — established automated regression testing
- Identified UI consistency and maintainability issues caused by the coexistence of a legacy design library and a new design system
- Documented component mappings and verification criteria, and applied gradual rollouts using feature flags
- Used Storybook and Playwright visual regression testing to fully remove legacy dependencies and establish a repeatable migration process
Reduced widget-installation support requests by 30% — built easy installation using DOM path extraction
- Identified recurring support requests because script insertion and display-position configuration required client engineering resources
- Implemented DOM path extraction for user-selected elements and dynamic widget insertion at the selected location
- Enabled widget installation without developer assistance, reducing installation-related support requests by 30%
Eliminated duplicate recommendations and unnecessary API calls — integrated Cafe24 cart data and improved recommendation logic
- Identified reduced recommendation relevance and unnecessary API costs from recommending items already in the cart and making calls in test environments
- Integrated the Cafe24 Front API SDK to retrieve cart items and exclude them from recommendations, with separate API-call conditions for test mode
- Improved widget display efficiency by eliminating duplicate cart-item recommendations and optimizing API calls for each execution environment