AI MVP · provider testing · launch gates
MAGIYA
An AI makeup-learning MVP: a user uploads a portrait, chooses a look and receives an AI-generated target image — with the goal of learning to recreate it.
Current status: MVP in development. The technical target-image slice passed testing; the full product is not yet ready for testers.
Problem
Virtual try-on shows a look but does not teach how to apply it. The product goal: help a person repeat a chosen makeup look on their own face.
What was built
- Product concept, user flows and specification for a portrait → target look → personal guidance journey.
- Comparison of AI image providers, including A/B rounds of generated results.
- Spend gates and tester gates before inviting first users.
- A readiness model that separates technical slice, product flow and real user outcome.
What it demonstrates
- Not letting a green technical test stand in for a delivered user outcome
- Explicit acceptance gates linked to the product promise
- Cost-aware experimentation with AI providers
My role
- Product concept and flows
- Provider comparison and scope decisions
- Tester gates, acceptance and spend control
Stack / methods
Evidence boundary
Backend and frontend were mostly AI/agent-assisted. Not a commercial launch; no users or revenue are claimed.