Beauty & Social
How Spyra Beauty Replaced Product Forms With an AI Scan and Passed 50K+ Installs
Techparser built AI product scanning and the social layer for Spyra Beauty, a live beauty app with 50K+ installs and 4.7 stars on Google Play.
- Client
- Spyra Beauty, Inc.
- Industry
- Beauty & Social
- Timeline
- December 2024 – October 2025
- Team
- 1 lead mobile engineer
Results
- 50K+
- Installs on Google Play as of September 2026
- 4.7★
- Google Play rating
- 5.0★
- App Store rating
The problem
Spyra Beauty is a digital makeup drawer. People save the products they own, review them and share the drawer with followers; the founder describes it as a place to scan and share the products you use. The friction was adding products. Typing a brand, product name, shade and category for every lipstick and serum takes a minute each, and a drawer with three items is not worth showing anyone. If adding products stayed slow, the drawers stayed thin and the feed stayed empty.
The app was already live on both stores when Techparser started work in December 2024, so features had to ship in steps alongside everything existing users relied on. The client also wanted the product to serve more than consumers: pageant and celebrity profiles with verified badges, giveaways to bring people back, and reports that tell brands who is using their products. All of it had to sit on a single Flutter codebase and Firebase backend without a rewrite.
What Techparser built
- AI product scanning: the user photographs a product, Google Cloud Vision reads the text on the packaging and OpenAI GPT-4o identifies the brand, product and details, which are filled into the form for the user to confirm.
- Receipt scanning: one photo of a receipt adds several products to the drawer at once.
- AI-generated art from a user's product collection, made to be shared, alongside QR sharing of a drawer.
- The social layer: a feed with saved posts, follows and followers, likes, comments, product reviews and ratings, friend activity, direct chat and push notifications.
- Pageant and celebrity profiles with verified crown badges, an all-stars ranking and giveaways.
- Brand reports exported to Excel for companies that want to understand who owns and reviews their products, plus an admin settings area for the Spyra team.
- Account and safety features: blocked users, account deletion flows, currency and language settings, and a forced-update check for old builds.
- Deep links so a shared drawer or product opens directly in the app, location search through Google Places on profiles, and push notifications sent from Cloud Functions through Firebase Cloud Messaging.
Decisions that mattered
Remove the typing
The core bet was that scanning a product or a receipt turns a minute of form-filling into a photo, and that full drawers are what make the social feed worth visiting. The scan pipeline pairs an OCR pass with a language model so that the model works from the text on the packaging rather than guessing from the image alone, and the result lands in an editable form. The user confirms rather than types, which keeps the fastest path and the most accurate path the same path.
New features without disrupting a live app
Spyra had users and reviews before this work began, so every feature shipped as an increment on the existing module structure: a module for scanning, another for receipts, another for pageant profiles, each with its own binding, controller and service. Firestore stayed the single source of truth, Cloud Functions handled server-side work such as notifications, and Firebase App Check and Crashlytics were added so new code paths could be watched in production. Releases went out through both stores as the client reviewed each step.
Build for the business as well as the user
A social app for beauty lovers has a second customer: the brands whose products fill the drawers. Techparser built the reporting side so Spyra could export who owns and reviews which products, and the pageant and celebrity verification so public figures could bring their audiences into the app. These features share models and services with the consumer side, which kept the product at one app rather than a consumer app plus a separate back office.
Watch it in production
Adding AI calls to a live app means adding new ways for it to fail: a model that times out, an image that cannot be read, a receipt in a currency the parser does not expect. Firebase Crashlytics and Analytics were wired so each new path could be measured, App Check limits backend access to genuine app builds, and a force-update check retires old versions once a breaking change ships. Problems surfaced in dashboards rather than in store reviews.
Outcome
As of September 2026 Spyra Beauty shows 50K+ installs and a 4.7-star rating on Google Play, and a 5.0-star rating on the App Store. The scanning, social and reporting features Techparser built are part of the shipping app.
The founder's review of the engagement: the work handled 'a lot of very complicated tasks' in a timely and efficient manner, and the Upwork contract that ran from December 2024 to October 2025 closed with a recommendation for Flutter and Firebase development.
“He helped handle a lot of very complicated tasks and did so in a timely and efficient manner. I would highly recommend…”
Questions about this project
- How does Spyra Beauty's product scanning work?
- The user photographs a product. Google Cloud Vision extracts the text on the packaging, then OpenAI GPT-4o identifies the brand, product and details from that text and the image. The result is filled into an editable form so the user confirms rather than types. A receipt variant of the same pipeline adds several products from one photo.
- What stack was Spyra Beauty built on?
- Flutter with for state management, and Firebase for Firestore, Auth, Storage, Cloud Messaging, Cloud Functions, App Check, Analytics and Crashlytics. Product recognition uses Google Cloud Vision and OpenAI GPT-4o, network calls go through Dio, and brand reports are generated as Excel files. Everything runs from one Flutter codebase on iOS and Android.
- Can Techparser add AI features to an app that is already live?
- Yes. Spyra Beauty was already in both stores when Techparser started. Features were added as separate modules, released in steps, and watched in production with Crashlytics and App Check so existing users were not disrupted. If you have a live Flutter or Firebase app and want scanning, recognition or generation features, that is the same engagement shape.
