POOJA Coconut Scan AI: Native App & B2B Marketplace

Pooja Scans AI: Revolutionizing B2B Agricultural Trade with Edge AI

1. Overview & Role

The Product: A centralized AI-powered B2B agricultural marketplace and admin dashboard. The platform enables users to instantly grade produce (coconuts) using AI, list them on a marketplace, and allows administrators to manage users, disputes, and subscriptions seamlessly. My Role: Lead Product Designer & Full-Stack Developer (End-to-End Execution) Platform: Native iOS & Android App, Web-based Admin Dashboard Tech Stack: React 19, Capacitor, Node.js/Express, TensorFlow Lite (.tflite)

2. The Challenge

The agricultural supply chain is plagued by manual, subjective grading processes and frequent disputes between buyers and sellers. Furthermore, grading often happens in rural areas with poor or zero internet connectivity.

The Business Goals:

1. Automate the grading process using AI to build trust between buyers and sellers.

2. Build a marketplace to facilitate these B2B transactions.

3. Create a robust Admin Dashboard to handle marketplace operations, payments, and disputes.

3. The Process (UX)

Before writing a single line of code or pushing pixels, I mapped out the complex ecosystem of our three core users: the Seller (Farmer/Trader), the Buyer, and the Platform Admin.

• Mapping the User Journey: I created user flows outlining how a seller would open the app, scan a batch of coconuts offline, sync the data when back online, and list it on the marketplace.

• Designing for the Environment: Because the mobile app would be used outdoors under bright sunlight, I prioritized high-contrast UI elements, large typography, and easily tappable touch targets to reduce cognitive load in the field.

• Admin Architecture: For the dashboard, the challenge was data density. I organized the navigation to separate user management, AI scan reports, financial operations (subscriptions/payments), and dispute resolution into digestible, scannable data tables.

4. The Solution (UI)

I developed a cohesive design system that translated perfectly from the high-density desktop admin view to the utilitarian mobile app.

• Real-Time AI Scanner: Designed a camera interface with an intuitive overlay, guiding the user on how to position the produce for the most accurate AI read.

• B2B Marketplace: Created a clean, trust-building marketplace UI where buyers can see the AI-verified grade, batch details, and seller ratings before purchasing.

• Centralized Admin Dashboard: Built a comprehensive web interface featuring data visualization for platform analytics, quick-action menus for resolving user disputes, and robust subscription management tools.

5. The Engineering (Dev)

Designing the experience was only half the battle; building it required bridging modern web technologies with native device capabilities.

• Cross-Platform Architecture: To move fast without sacrificing quality, I built the core application in React 19. I then utilized Capacitor to wrap the web application into native iOS and Android builds, allowing for a single codebase to serve multiple platforms.

• Offline Edge AI Integration: The most significant technical challenge was the AI scanner. I integrated a .tflite (TensorFlow Lite) model directly into the native app layer. This allowed the app to access the device's camera via native bridges and process the image grading locally on the edge—meaning users could grade hundreds of coconuts flawlessly without an internet connection.

• Full-Stack Syncing: I engineered a custom Node.js backend that handles local data caching. Once a user regains internet access, the app silently syncs the offline AI scan reports and marketplace listings to the central database, instantly updating the Admin Dashboard.

Key Screens:

• AI Scan Flow (Image → Video → Audio → Result)

• Marketplace

• Bargain/Negotiation

• Order Tracking

• Seller Dashboard

• Admin Dashboard

• Subscription Plans (Guest, Gold, Diamond)

  

6. The Outcome

By owning both the design and the development, I ensured that the initial UX vision wasn't compromised by technical limitations.

• Successful Cross-Platform Launch: Delivered a highly performant native mobile app and a complex web dashboard simultaneously.

• Zero-Latency Grading: By pushing the AI to the edge (.tflite), we achieved instant grading feedback for users, completely bypassing server latency and connectivity issues.

• Streamlined Operations: The centralized admin dashboard reduced the time required to manage marketplace disputes and monitor AI scan accuracies, giving the business a clear path to scale.

"This project proved the value of a hybrid approach: designing with technical constraints in mind, and coding with an unwavering focus on the user experience."

Reference video:
https://drive.google.com/drive/folders/13VfbdewlTAiIblABGBuJryQOKtfHK8Uo?usp=sharing

Analyze the reference video and redesign the experience with better UI/UX, smoother interactions, and production-ready mobile design.