파머 랭크 AI: 생성 AI를 사용하여 농부와 올바른 구매자를 연결

Introduction

Hello, I am Shaik Inzamam, a B.Tech Computer Science student and aspiring AI engineer from India.

I enjoy building practical AI solutions using Generative AI, large language models, backend development, cloud technologies, and automation tools. I am especially interested in creating solutions that address real problems faced by local communities.

For the Gen AI Academy APAC – Meet the Builders initiative, I am sharing my project:

Farmer Rank AI

Farmer Rank AI is a Generative AI-powered agricultural marketplace assistant that helps buyers find and connect with suitable farmers based on crop type, quantity, quality, price, and location.

The goal of the project is to make agricultural trading more accessible, transparent, and efficient for farmers and buyers in India.

The Local Problem

Agriculture plays an important role in India, but many farmers still face challenges when trying to connect directly with genuine buyers.

Farmers may have quality produce available, but they often lack visibility and access to the right market. Buyers also face difficulties identifying farmers who can meet their exact requirements.

For example, a buyer may need:

500 kilograms of tomatoes
Grade A quality
A maximum price of ₹20 per kilogram
A farmer located near Bengaluru
Immediate availability

Finding the right farmer manually can take a lot of time. Buyers may need to contact several farmers individually before finding a suitable match.

At the same time, smaller farmers may be ignored because they do not have access to advanced digital platforms or strong marketing networks.

Some of the major problems include:

Difficulty connecting farmers directly with buyers
Limited visibility for small and regional farmers
Time-consuming manual search
Lack of intelligent farmer ranking
Incomplete information about crop quantity and quality
Poor location-based matching
Lack of transparency in recommendations
Communication gaps between farmers and buyers
The Idea Behind Farmer Rank AI

Farmer Rank AI acts as an intelligent bridge between agricultural buyers and farmers.

Instead of manually selecting multiple filters, a buyer can describe their requirement in normal language.

For example:

I need 500 kg of Grade A tomatoes near Bengaluru for under ₹20 per kg.

The application understands the request, extracts the important information, searches available farmer listings, and recommends the most suitable farmers.

The platform does not simply return a long list of results. It ranks farmers according to how closely they match the buyer’s needs.

It also explains why each farmer was recommended.

Example Buyer Request

A buyer can enter:

Find farmers who can supply 500 kg of Grade A tomatoes near Bengaluru for less than ₹20 per kilogram.

The application extracts structured information such as:

{
“crop”: “tomato”,
“quantity_kg”: 500,
“quality_grade”: “A”,
“preferred_location”: “Bengaluru”,
“maximum_price_per_kg”: 20
}

The system then searches the available farmer data and evaluates each farmer using multiple conditions.

These conditions include:

Crop availability
Quantity available
Quality grade
Price per kilogram
Location
Distance from the buyer
Farmer reliability
Previous buyer preferences
How Farmer Rank AI Works

The application follows a structured AI workflow.

Step 1: Buyer Enters a Requirement

The buyer enters a crop requirement using natural language.

The buyer does not need to understand technical filters or database fields.

Example:

I need 1,000 kg of onions near Hyderabad within ₹25 per kilogram.

Step 2: Gemini Understands the Request

Google Gemini processes the request and identifies important information such as:

Crop name
Required quantity
Quality grade
Location
Maximum budget
Buyer intent
Step 3: Structured Filters Are Created

The natural-language request is converted into structured search filters.

This makes it easier for the backend to search available farmer listings.

Step 4: Matching Farmers Are Retrieved

The system searches farmer records based on the extracted requirements.

Farmers who do not meet the basic conditions are removed.

For example, a farmer may be excluded when:

The crop does not match
The available quantity is too low
The price exceeds the buyer’s budget
The farmer is too far from the preferred location
The required quality grade is unavailable
Step 5: Farmers Are Ranked

The remaining farmers receive a ranking score.

The ranking system considers:

Crop match
Quantity match
Quality match
Price match
Location relevance
Distance
Farmer availability
Buyer preferences
Step 6: Gemini Generates Explanations

Gemini generates a simple explanation for each recommendation.

Example:

This farmer is recommended because they have 600 kg of Grade A tomatoes available at ₹18 per kilogram and are located close to Bengaluru.

Step 7: Buyer Reviews the Results

The buyer receives a ranked list of farmers with:

Farmer name
Crop
Available quantity
Quality grade
Price
Location
Matching score
Recommendation reason
Contact option
Step 8: Buyer Connects With the Farmer

After reviewing the results, the buyer can contact the selected farmer and continue the discussion.

Google AI and Google Cloud Technologies

Google technologies play an important role in the architecture of Farmer Rank AI.

1. Google Gemini API

Google Gemini is the main Generative AI layer of the project.

It is used to understand buyer requirements written in natural language.

Gemini helps the application perform tasks such as:

Detecting buyer intent
Extracting crop names
Extracting required quantities
Understanding quality grades
Identifying locations
Extracting budget constraints
Converting text into structured data
Generating farmer recommendation explanations
Summarizing matching results
Supporting multilingual requests
Producing easy-to-understand responses

Gemini makes the application more natural because buyers can communicate with the system like they would communicate with another person.

1. Vertex AI

Vertex AI provides a managed platform for working with Gemini models on Google Cloud.

Vertex AI can help Farmer Rank AI:

Access Gemini securely
Manage AI model requests
Configure generation settings
Handle production workloads
Monitor AI usage
Evaluate model responses
Improve prompt quality
Scale AI operations

Vertex AI can also support future improvements such as prompt evaluation, model monitoring, and AI response testing.

1. Google Cloud Run

The backend of Farmer Rank AI can be deployed using Google Cloud Run.

Cloud Run allows the application to run as a containerized service without manually managing servers.

Cloud Run is useful because it provides:

Automatic scaling
Container-based deployment
Secure API hosting
Reduced infrastructure management
Faster deployment
Support for backend services
Pay-per-use infrastructure

When a buyer submits a query, the backend running on Cloud Run receives the request, communicates with Gemini, searches farmer data, and returns ranked recommendations.

1. Google Cloud Firestore

Google Cloud Firestore can store structured information used by the application.

This includes:

Farmer profiles
Farmer names
Crop listings
Crop quantity
Crop quality
Price per kilogram
Farmer location
Contact information
Buyer preferences
Search history
Recommendation history
Crop availability status

Firestore supports real-time updates, which can be useful when farmers update crop quantity, price, or availability.

1. Google Cloud Storage

Google Cloud Storage can be used to store larger files that should not be saved directly in the database.

Examples include:

Crop images
Farmer profile images
Farmer verification documents
Quality certificates
Dataset files
Generated reports
Application screenshots
Demo assets

The database can store the file reference, while the actual file remains securely stored in Cloud Storage.

1. Google Maps Platform

Google Maps Platform can improve location-aware farmer recommendations.

It can help the application:

Convert addresses into coordinates
Calculate distances
Display farmer locations
Identify nearby farmers
Estimate travel distance
Support regional searches
Improve logistics planning

For example, when a buyer requests tomatoes near Bengaluru, the system can prioritize farmers located closer to the buyer.

Location information can become an important part of the ranking score.

1. Google Cloud Secret Manager

Farmer Rank AI uses several sensitive credentials, such as AI keys, database credentials, and service configuration.

Google Cloud Secret Manager can securely store:

Gemini credentials
Database connection details
Authentication secrets
External API keys
Application configuration values

This prevents sensitive keys from being written directly into the project source code.

1. Google Artifact Registry

Farmer Rank AI uses Docker for containerization.

The backend Docker image can be stored in Google Artifact Registry before deployment.

The deployment process can follow these steps:

Build the application
Create the Docker image
Push the image to Artifact Registry
Deploy the image to Cloud Run
Make the backend API available to the frontend

Artifact Registry provides a secure location for storing and managing container images.

1. Google Cloud Build

Google Cloud Build can automate the project build and deployment workflow.

When new code is pushed, Cloud Build can:

Install project dependencies
Run build commands
Build the Docker image
Push the image to Artifact Registry
Deploy the application to Cloud Run

This makes deployment faster and reduces manual work.

1. Cloud Logging

Cloud Logging can record important application events.

Examples include:

Buyer queries
Gemini requests
Gemini response errors
Farmer search results
Ranking workflow activity
Database errors
API failures
Safety validation events

Logging makes it easier to debug the application and understand how users interact with the system.

1. Cloud Monitoring

Cloud Monitoring can track the health and performance of the application.

It can monitor:

Number of requests
API response time
Error rate
Cloud Run performance
Resource usage
Service availability
Failed AI requests

Monitoring is important for maintaining a reliable production application.

1. Firebase Authentication

Firebase Authentication can be used to manage secure user access.

It can support:

Farmer registration
Buyer registration
Email login
Google login
User authentication
Role-based access
Secure sessions

Farmers and buyers can have separate profiles and different application permissions.

1. Firebase Hosting

The frontend application can be hosted using Firebase Hosting.

Firebase Hosting can provide:

Fast web hosting
HTTPS support
Global content delivery
Easy deployment
Integration with Firebase services

The Next.js frontend can communicate with the backend deployed on Cloud Run.

Google-Powered Workflow

The complete Google-powered application workflow can operate as follows:

A buyer opens the web application hosted using Firebase Hosting.
The buyer signs in through Firebase Authentication.
The buyer enters an agricultural requirement in natural language.
The frontend sends the query to the backend deployed on Cloud Run.
The backend sends the buyer query to Gemini through Vertex AI.
Gemini identifies the buyer’s intent and converts the request into structured filters.
Farmer listings are retrieved from Firestore.
Google Maps Platform calculates location relevance and distance.
The ranking workflow evaluates crop, quantity, grade, price, and proximity.
Gemini generates a clear explanation for every recommendation.
Crop images and farmer documents are retrieved from Cloud Storage when required.
Ranked farmer results are returned to the buyer.
Cloud Logging records important workflow events.
Cloud Monitoring tracks service performance.
Secret Manager protects sensitive credentials.
Cloud Build and Artifact Registry support deployment updates.
System Architecture

Farmer Rank AI uses a modular architecture so that each component has a clear responsibility.

Buyer
|
v
Next.js Web Application
|
v
Firebase Authentication
|
v
Backend API on Google Cloud Run
|
+————————–+
| |
v v
Gemini through Vertex AI Firestore
| |
v v
Intent Extraction Farmer Records
| |
+————-+————+
|
v
Google Maps Platform
|
v
Farmer Ranking Engine
|
v
Gemini Explanation Generator
|
v
Ranked Farmer Results

Other Google Cloud services support the application:

Cloud Storage -> Crop images and documents

Secret Manager -> API keys and credentials

Artifact Registry -> Docker images

Cloud Build -> Automated deployment

Cloud Logging -> Application logs

Cloud Monitoring -> Performance monitoring
Complete Technology Stack
Google AI and Cloud Stack
Google Gemini API
Vertex AI
Google Cloud Run
Google Cloud Firestore
Google Cloud Storage
Google Maps Platform
Google Cloud Secret Manager
Google Artifact Registry
Google Cloud Build
Google Cloud Logging
Google Cloud Monitoring
Firebase Authentication
Firebase Hosting
Application Development Stack
Next.js
Node.js
TypeScript
Express.js
Mastra
Docker
Git
GitHub
Data and Retrieval Stack
PostgreSQL
Qdrant
Vector-based retrieval
Structured farmer filtering
AI-powered ranking workflow