Skip to content
Tamar Mkurnalidze, MA MSc
Home

Case study · AI-assisted product

Veylora

An AI-assisted personal styling and reflection product — designed and built end-to-end as a personal software project.

Role
Solo designer & builder
Timeframe
2024 – ongoing
Status
Private preview

Problem

Personal styling apps typically stop at outfit suggestions and ignore the emotional context around getting dressed. I wanted to explore whether a small, opinionated AI-assisted product could combine practical wardrobe decisions with a warmer, gentler tone — without over-promising and without collecting sensitive data.

My role

Solo end-to-end: concept, research, product architecture, UX flows, prompt engineering, front-end build, backend integration, private preview and content. AI-assisted development throughout, with clear scope discipline.

Research

  • Comparative review of AI stylist products and personal colour analysis frameworks.
  • Interviews and observation of how people actually use their wardrobes on ordinary mornings.
  • Literature scan on decision fatigue and small-choice design in consumer apps.

Design

  • Editorial visual system: warm porcelain background, deep espresso ink, champagne accent.
  • Single-column, mobile-first flows optimised for one-handed use in under 10 seconds.
  • Distinct tones for utility features (styling) and reflective features (companion), avoiding a single flat brand voice.

Technology

  • TanStack Start on Cloudflare Workers for the app shell and server functions.
  • Supabase for authentication, row-level security and structured data.
  • Lovable AI Gateway for model access, with structured system prompts and JSON-mode outputs for wardrobe tagging.
  • Tailwind CSS v4 with a semantic token palette.

Lessons learned

  • AI-assisted development is fast — scope discipline still wins.
  • The smallest testable version of an interaction teaches more than the most complete spec.
  • Voice and tone are product decisions, not decoration.
  • Restraint on features is what makes a small product feel serious.

Future improvements

  • Deeper wardrobe analytics — cost per wear, colour balance over time, gap analysis.
  • Structured content library (essays, notes) rather than only interactive features.
  • Formal accessibility audit and colour-contrast tuning against WCAG 2.2 AA.
  • A public read-only mode that lets visitors experience the product without signing in.