Case study · MVP we built

Naina: an AI coaching app, from first screenshot to paid pass

Naina is an AI coaching app MVP we built end to end. A user uploads screenshots of their dating or matrimony profile, gets a free score and the top fixes without making an account, then signs in with an emailed one-time code to save their history and can buy a Coach Pass for the full plan.

Type
AI coaching app, built as an MVP
Flow
Upload, free score, sign-in, paid pass
Sign-in
Email one-time code, no password
Payments
Stripe checkout, with a mock mode for testing
Stack
Next.js, React, TypeScript
Quality
Rule tests, scoring evals and an end-to-end money-path test

The problem

People tune a dating or matrimony profile by guessing. Friends are kind, advice online is generic, and nobody can say which photo or which line of the bio is holding the profile back.

An MVP for that has to prove one thing: that a stranger will upload a screenshot, believe the answer and come back for more. Everything else can wait, and AI chat about personal doubts was deliberately left for later.

What we built

The product is one flow, built so each step earns the next one.

  • Upload without an account. A visitor drops in screenshots of their profile and gets a free score with the top three fixes before being asked to sign up.
  • Email one-time-code sign-in. A six-digit code sent by email saves history to the user's account, so there is no password to forget.
  • A paid Coach Pass. Stripe checkout unlocks the full action plan, a bio rewrite, a presentation plan and voice practice, and the unlock is verified on the server before it appears.
  • A dashboard. History, the looks plan and voice practice sit in one place after sign-in.
  • Prices set per market. Seven currencies carry fixed local price points instead of live conversions, so a customer never sees an odd amount that moves underneath them.

How the AI is kept honest

The part of an AI product that decides whether people trust it is how its answers are checked. We split the work so that rules do what rules do well and a model only does what needs wording.

  • A fixed rule set does the scoring. Scores come from tested rules across weighted categories, so the same screenshots give the same score every time, with no per-review model cost.
  • AI adds the personal touch, with guardrails. Where a model writes, such as a bio rewrite or a personal note, its answer must match a strict format, is screened for safety and is cached. If one provider is rate-limited or not configured, the next one takes over.
  • Evals as a regression check. A command runs the engine on sample profiles, such as strong photos with a weak bio, and prints the scores, so a tuning change that moves them is caught straight away.
  • Tests on the path that earns money. An end-to-end test walks the whole route from free score to sign-in to Coach Pass to premium content in a mock-payment mode, so no card is charged while it runs.

What it shows about how we would build yours

Naina is a small product with a complete business loop: free to try, paid to unlock and tested from the first click to the unlocked page. That is the shape of work we take on for founders and for businesses adding AI to a product.

  • An AI feature with a business model. The free step exists to earn the paid one, and the payment path is tested end to end.
  • AI you can check. Rules where rules work, models where wording matters, and tests for both.
  • Small first, ready to grow. The MVP covers the one flow that has to work, on mainstream tools, so a larger team can pick it up later.

Related services

Want something like this built for your business?

No call needed. You get the approach, a price range and a timeline in writing, free.