Eoin Cronin

Lead Frontend & UX EngineerCork, Ireland

I turn design systems into production code.

Eight years shipping Angular and React, now making AI-generated UI meet the same standard.

Work

  1. 01Architecture

    A design system the whole org adopted

    I led an organisation-wide Figma design system built on Atomic Design, mirrored its variables as typed tokens in code, and linked Figma components to their Angular twins with Code Connect, so the Storybook library and Figma stay in step.

    Figma → Angular Storybook

    Select a stage

    Example: Variables

    Figma
    color/action/primary
    Token
    --color-action-primary: #ff4f00;
    Type
    type ColorToken = "action-primary" | "action-secondary";
    Scale
    8 production apps
    Sync
    1:1 Figma Code Connect
    System
    Typed atomic tokens

  2. 02Workflow

    AI-native design-to-code, with guardrails

    I connected Figma to Cursor through MCP and the Figma API. Because the design system is linked with Code Connect, Cursor builds with the real components instead of guessing, and prompt and review standards hold AI-generated UI to the same bar as hand-written code.

    Design to production

    Select a step

    + Code Connect feeds Cursor the design system's real components

    Rule: Guardrails

    Design-system components only, tokens instead of raw values, reviewed to the same bar as hand-written code.

    Post-release defects

    Post-release frontend defects fell by 30 percent.

    Defect reduction
    −30% post-release
    Pipeline
    Figma MCP → Cursor
    Standard
    Design-system components only

  3. 03Research

    Where a university app fails part-time students

    UX research & service design study, Postgraduate Diploma in Digital Design (NFQ Level 9), TU Dublin

    A design research study of TU Dublin's new student app, from the side of part-time postgraduates who study online. I ran a service safari and three moderated interviews, benchmarked Coursera, Airbnb and Google Maps, and synthesised the results into four findings and six prioritised recommendations, each with a success measure.

    Methods

    Select a step

    Synthesis produced

    Recommendations

    Delivered 6 prioritised recommendations, each with a success measure, such as 3 taps or fewer to contact a lecturer, down from 13.

    Taps to reach a lecturer

    TU Dublin app
    13
    Airbnb
    2
    Target
    ≤ 3
    Friction target
    13 taps → ≤ 3
    Method
    Service safari + benchmarking
    Output
    6 metric-backed recommendations