MIRRR

Most beauty apps were built for the West. MIRRR was built for the rest. Multicultural Intelligence for Real-time Recommendation & Reflection powered by Qloo's cultural taste graph.

Comment

Inspiration

I kept noticing the same gap: beauty AI is everywhere, but it was built for one face. Western, fair-toned, Eurocentric. If you're South Asian, Middle Eastern, West African, Latina, or East Asian, every app tells you to "find your closest match." There is no closest match. There's your match.

The insight that broke it open: culture is a better signal than demographics. A woman who grew up on Afrobeats, jollof rice, and Nollywood films has a different beauty aesthetic than a woman with the same skin tone who grew up on K-dramas and matcha. Skin tone tells you nothing. Cultural taste tells you everything.

Qloo has the cultural taste graph. MIRRR puts it to work.

What it does

MIRRR is a Qloo-powered ethnic beauty discovery agent. You take a 5-question Cultural DNA quiz — not about your skin tone, but about your world: the music you love, the food you eat, the aesthetics that move you. The MirrrAgent resolves those signals through Qloo's Taste AI graph (250M+ entities), builds a real cultural taste profile, and returns:

  • ShadeFinder — foundation, concealer, and lip shades calibrated to your actual undertone and depth
  • WigMatchmaker — protective styles and textures aligned with your cultural aesthetic
  • LookEngine — full beauty looks anchored to your Qloo taste profile, not a demographic checkbox

No proxies. No guessing. Real cultural intelligence.

How we built it

Pure Node.js — zero npm packages, zero frameworks. Every module hand-built using only Node.js built-ins (http, fs, url, path). ~75,000 lines across 27 files, all ESM modules.

The core is the MirrrAgent — a 6-step agentic pipeline:

  1. Perceive — parse quiz responses and stated preferences
  2. Resolve — call Qloo's entity resolution API to anchor inputs to real cultural entities
  3. Profile — build a cross-domain taste profile via Qloo's affinity graph
  4. Reason — score against 10 Cultural DNA communities (Afro-Diasporic, Desi, MENA, East Asian, Latinx, and more)
  5. Recommend — run ShadeFinder, WigMatchmaker, and LookEngine in parallel
  6. Narrate — generate a personalized beauty story explaining why these picks fit you

Each step depends on the output of the previous. Qloo powers steps 2 and 3 — remove it and the entire downstream collapses into generic guessing.

Challenges we ran into

  • Building a complete agentic pipeline with zero npm dependencies — every utility, every router, every parser written from scratch
  • Designing a Cultural DNA scoring system that is culturally precise without being reductive or stereotyping
  • Making Qloo meaningful at every pipeline step, not just a decorative API call at the end
  • Solo build, 48 hours, $0 infrastructure (Render free tier) ## Accomplishments that we're proud of
  • A genuinely working 6-step agentic pipeline that would break without Qloo — not a wrapper, a real dependency
  • 10-community Cultural DNA Engine that scores cross-domain taste signals into actionable beauty profiles
  • Three specialized engines (ShadeFinder, WigMatchmaker, LookEngine) each using Qloo differently
  • ~75,000 lines of pure Node.js shipped solo in 48 hours with no framework scaffolding
  • A live deployed app that any judge can use right now: https://mirrr-upbn.onrender.com ## What we learned Qloo's cross-domain affinity is the thing no demographic checkbox can replicate. The same API that knows someone who loves Kendrick Lamar also leans toward certain fashion aesthetics — that's the signal beauty AI has been missing. Cultural taste is a richer input than any form field about skin tone or hair type. I also learned that the absence of npm doesn't mean the absence of capability. It means you understand exactly what you built.

What's next for MIRRR

The global South beauty market — India, Nigeria, Brazil, Indonesia, the Gulf — is massively underserved by existing beauty tech. MIRRR is a founding thesis, not just a hackathon project.

Next: integrate live product inventory from GRWM Collective, SheaMoisture, Kay Beauty, and Huda Beauty. Add Qloo-powered community discovery so users find others who share their aesthetic — not their ZIP code. Build the brand partnership layer so culturally-aligned beauty brands can reach audiences Sephora's algorithm has never found.

Most beauty apps were built for the West. MIRRR was built for the rest.

Built With

Share this project:

Updates

Submission history