Challenge
Manual grading couldn't scale
Every diamond assessment required a certified expert evaluating each stone individually. As volume grew, the process created a bottleneck — more demand meant more specialists, not faster throughput.
Case Study
Yehuda
AI-powered diamond detection — built to scale precision grading without scaling headcount.
2025Luxury Fashion and Jewelry
Client
Yehuda
Sector
Luxury Jewelry
Services
AI Development, Mobile App
Year
2025
Live
App Store & Google Play
Overview
Yehuda
Yehuda is a fine jewelry and diamond specialist operating across multiple markets, with a client base that expects the same precision in digital service as they do in the physical product. The core challenge was manual: diamond assessment and grading required expert human evaluation for every stone — a process that couldn't scale with demand without proportionally scaling headcount.
Mejix was brought in to design and build a custom AI-powered mobile application that could automate the diamond detection and grading process — reducing the time and cost per assessment while maintaining the accuracy that Yehuda's clients and reputation depend on.
What this project delivered
A production-ready AI mobile product built for precision operations, faster assessment workflows, and auditable grading history.
Challenges
We built Yehuda's AI Diamond Detection Platform
Automating precision assessment in a field where accuracy is non-negotiable required solving four distinct technical and operational problems simultaneously.
Challenge
Accuracy had to match human expert levels
The AI system couldn't just be fast — it had to be accurate to the standard that a professional grader would produce. Any drop in grading quality would undermine client trust in the platform immediately.
Challenge
The app had to work across geographies
Yehuda's clients and assessors are geographically dispersed. The solution needed to work reliably across different devices, lighting conditions, and network environments — not just in a controlled lab setting.
Challenge
Integration with existing workflows
The AI detection system needed to integrate into Yehuda's existing operational workflow — not replace it entirely. Assessors needed to be able to review, override, and act on AI outputs within the same tool they already used.
Approach
How we turned the concept into a working product
We approached the project as both a product and an AI systems challenge. The mobile experience had to be simple enough for day-to-day use in the field, while the inference pipeline had to standardise image quality, handle inconsistent conditions, and return results fast enough to fit seamlessly into Yehuda's grading workflow.
That meant pairing a cross-platform React Native client with a purpose-built computer vision backend, then designing the surrounding operational layer — history, review notes, overrides, and assessment logs — so the app could support real-world grading operations rather than just demo AI output.
Solutions
A custom AI detection system built for precision at scale
We designed and built a React Native mobile application with a custom computer vision pipeline — enabling automated diamond detection and grading at a consistency and speed that manual assessment cannot match.
AI-Powered Diamond Detection
A custom computer vision model trained to identify, classify, and grade diamonds from mobile camera input — delivering assessment results that match the accuracy of a certified human grader, in a fraction of the time.
Cross-Platform Mobile App
Built in React Native for iOS and Android, the app works across device types and under real-world conditions — inconsistent lighting, varied camera quality, and unreliable network connectivity — without sacrificing accuracy.
Assessment History & Reporting
Every grading session is logged with the stone details, AI assessment output, assessor review notes, and timestamp — giving Yehuda a complete, auditable record of every diamond that has passed through the platform.
Technology
Tech stack chosen for mobile AI performance
Built on a modern, production-ready stack selected specifically for the performance and accuracy requirements of AI-powered mobile computer vision.
React Native
Cross-platform mobile development for iOS and Android — a single codebase delivering native performance on both platforms, with access to device camera APIs required for the vision pipeline.
OpenAI Vision API
The AI backbone of the diamond detection system — custom-prompted and fine-tuned for the specific visual classification task, integrated via a Python backend optimised for low-latency mobile inference.
Python / FastAPI
The AI inference layer runs on a Python FastAPI backend — handling image preprocessing, model inference, result formatting, and the API responses consumed by the React Native client.
Supabase
PostgreSQL-backed database and auth layer — storing assessment records, user accounts, stone history, and AI output logs with row-level security and real-time sync to the mobile client.
AWS
Cloud infrastructure for the AI inference backend — EC2 for the FastAPI service, S3 for image storage, and CloudFront for global CDN delivery of static assets and model artefacts.
Custom CV Pipeline
A purpose-built computer vision preprocessing pipeline — normalising input images for lighting, angle, and resolution before passing to the model, ensuring consistent inference quality regardless of device or shooting conditions.
Results
From launch to measurable impact
Yehuda's AI detection platform delivered verified operational results from launch without relying on inflated vanity metrics.
5,000+
Patients diagnosed from launch
Verified impact from the live Yehuda deployment.
On time
Delivered on time and within budget
Confirmed in the published Yehuda testimonial.
2 platforms
Live on iOS and Android
Cross-platform delivery with a single production product.
Certifications & Recognitions
Reviews & Certifications
Our clients praise us for
our great results
Mejix has successfully created and launched a functional mobile app, which allows the client to reach and diagnose over 5,000 patients. Despite geographically dispersed clients, the team is accommodating and communicative. They've worked hard to deliver the project on time and within the budget.
Candice Wu
Head of Design, Yehuda
Why Mejix
Frequently asked questions
Answers to the questions buyers usually ask after seeing this case study.
The Yehuda AI diamond detection application was designed, built, tested, and launched within the project timeline — delivered on time and within budget. Contact us for detailed timeline information relevant to a similar AI mobile application project.
Case Studies
You might like
Finance and Banking
Dish by Metro
We support DISH by Metro's Dish Pay platform with dedicated cross-functional teams delivering end-to-end product development — a high-volume digital payment solution serving the US restaurant market.
2025
Finance and Banking
BRSK
We provide BRSK with a dedicated engineering team through a long-term staff augmentation model — senior developers embedded in their delivery process, supporting platform growth with scalable, high-performance engineering.
2025
Retail and Services
Verstolo
We built Verstolo's fine jewelry ecommerce store end-to-end on Shopify Plus — bespoke storefront design and development built to match the brand's premium positioning.
2025
How can we help?
Whether you're building a new product, scaling your platform, or integrating AI — our team is ready to deliver.