Product Recommendation System Development for Ecommerce: Clime

Industry	Platform	Solution We Provided
eCommerce	Web Application	AI-powered Product Recommendation Engine

Project Overview

Our client represents an innovative eCommerce entrepreneur focused on sustainable living solutions. Clime, founded with a vision to make eco-friendly shopping accessible to Gen Z consumers, aims to create a centralized platform for environmentally conscious product discovery. The client also wanted to enable smart product discovery through an AI-based recommendation engine.

With a strong commitment to promoting sustainable alternatives in household and personal care products, our client recognized the growing demand among younger consumers. These consumers seek products that are not only effective but also environmentally friendly, safe for sensitive skin, and suitable for babies.

What makes this venture different? Unlike traditional eCommerce platforms, Clime focuses exclusively on curating products that are economically sustainable and environmentally friendly, ensuring every recommendation aligns with its users’ values and health considerations. Our client wanted to aggregate these specialized products and showcase them on their own platform.

It’s time to turn our focus to the challenges our client faced when they first approached us.

Challenges Our Client Faced

The sustainable eCommerce landscape presented unique challenges that our client needed to overcome to create a successful product recommendation platform:

1. Lack of guaranteed sustainable product discovery: Multiple eCommerce websites and manufacturer stores existed, but none provided guarantees for sustainable or environmentally friendly products, especially those safe for sensitive skin and babies.
2. Competitive market validation: Perplexity was already providing “shop with AI” features, so our client needed to validate their sustainable-focused approach through a proof-of-concept before full development.
3. Brand repetition in recommendations: Traditional recommendation systems showed the same top products repeatedly. For example, if a database contained 120 toilet cleaners, users would always see the same top 3 results, creating an appearance of promoting certain brands over others.
4. Limited product comparison and follow-up: Existing platforms didn’t provide comprehensive feature comparisons for eco-friendly products or contextual follow-up capabilities for deeper product exploration.

Building upon these challenges, our client approached us with a comprehensive set of requirements for their proof-of-concept solution.

Client Requirements 

Our client wanted us to build an AI-powered recommendation system that would serve as a proof-of-concept to validate their sustainable eCommerce store idea.

Some of the most notable requirements they highlighted were:

Now, allow us to give you a glimpse into the solution we delivered to address their requirements. Our AI software development services can help you build similar intelligent eCommerce solutions.

Our Solution Approach

We designated one expert full-stack developer to deliver this proof-of-concept solution, focusing on creating a robust foundation for the client’s sustainable eCommerce vision.

As a leading AI development company, we at Space-O Technologies AI prioritize our clients’ vision and consider it our duty to deliver solutions that validate business concepts while providing scalable technical architecture for future development. To learn more about our approach, take a look at the process our developer followed:

1. Discovery and Validation Phase

Our developer worked closely with the client to understand the specific requirements for sustainable product curation and Gen Z user preferences. We analyzed the competitive landscape, particularly studying existing “shop with AI” approaches to adapt them for sustainable product discovery.

2. AI Architecture Design

We designed a hybrid recommendation system that addresses the core challenge of brand variety. Instead of simply showing the top three vector similarity results, our system identifies three different brands and selects their best products, incorporating controlled randomization to ensure diverse recommendations.

3. Vector Database Implementation

Our team created sophisticated product embeddings by generating vectors from uploaded product information and automatically creating additional meta fields. This approach ensures comprehensive product representation in the vector database for accurate similarity matching.

4. Hybrid Intent Classification Development

We implemented a three-part intent classification system that uses AI, pattern matching, and direct database queries to understand user intent and provide appropriate responses based on conversation context.

By following this comprehensive approach, we were able to deliver a proof-of-concept solution that validated our client’s hypothesis while providing a solid foundation for future enhancements.

Technology Stack

It’s time we delve deeper and learn about the tech stack our team used for creating this solution:

Technology	Use
React with TypeScript	User interface development
NestJS (Node.js) with TypeScript	Server-side development
PostgreSQL with Sequelize ORM	Primary database management with vector extensions (pgvector)
Tailwind CSS with Radix UI	UI styling and component library
OpenAI API	Natural language processing and conversation
Auth0 (Passwordless), JWT tokens	User authentication system
Vector database	Similarity search for product recommendations

Using the mentioned technologies, we built a proof-of-concept AI-powered recommendation engine that validated our client’s approach to sustainable product discovery. It also demonstrated the feasibility of brand diversification and conversational user interaction.

Solution Overview

After comprehensive development and testing, you might wonder what makes this recommendation system different, right? Let us break it down for you.

Our solution creates an intelligent chatbot where users discover sustainable products through natural language conversations. Users type requests like “I am looking for a refrigerator” and receive three curated recommendations from different brands, eliminating the repetitive results of traditional systems.

The platform generates product vectors from uploaded information and automatically creates meta fields for accurate matching. It provides five follow-up questions after each recommendation, enabling deeper exploration through continued conversation similar to modern AI assistants.

The example below illustrates how Clime’s recommendation engine presents comprehensive product comparisons, displaying detailed attributes across multiple sustainable alternatives. This helps users make informed purchasing decisions.

We will now discuss how the AI-powered recommendation system addresses the specific challenges our client faced in sustainable product discovery.

Impact of Our Solutions

The recommendation system we delivered successfully addressed the key challenges our client identified in sustainable product discovery:

1. Eliminated Brand Repetition

Our hybrid algorithm solved the core problem of showing the same products repeatedly. Users now receive varied recommendations from three different brands instead of seeing identical results every time.

2. Streamlined Sustainable Product Discovery

The platform provides a centralized solution for discovering sustainable, environmentally friendly products specifically curated for safety and environmental impact.

3. Enhanced User Interaction Through Follow-up Questions

The system generates contextual follow-up questions that keep users engaged and help them explore product options more thoroughly.

4. Intelligent Context Understanding

The hybrid intent classifier successfully distinguishes between new product searches and questions about previously recommended products, enabling natural conversation flow.

5. Efficient Product Management

The admin panel allows manual product curation while automatically generating vectors and meta fields, creating a scalable foundation for the sustainable product database.

This proof-of-concept system validated our client’s hypothesis and provided a solid foundation for future development phases, including potential automation through scraping and enhanced AI capabilities.

Now, let’s proceed and take a brief look at some of the key features that the AI-powered recommendation system offers to both administrators and end users.

Key Features by Stakeholder

1. End Users

• Document Management: This feature allows the admin to create, update, and import multiple document types. It enables them to sync, rename, archive, and delete documents. Additionally, they can add labels and sub-labels to documents for better sectional distinctions.
• User Management: This feature allows the admin to add, manage, and delete user roles. They can add new users by adding simple details like their name, email address, password, mobile number, and project. 
• Role Management: The admin can assign specific roles to the added users, such as super admin and admin. They can also grant them certain permissions to access various functionalities, such as document management, user management, and profile management. 
• Profile Management : Admin can update their profile details, such as name, email ID, organization name, number, and password. This provision makes it easier for them to keep their credentials up-to-date, especially during interstate transfers.
• Chatbot: This chatbot will be available for every document processed and analyzed on the platform. The admin will also have a filter option to customize the chatbot based on two categories: labels and documents. 
• Share Chatbot: The admin can share chatbots with followers by generating shareable links. They have two options when sharing chatbots: private or public, thereby generating universal links and customizable private links. 
• AI-Powered Product Search (Text + Images): Users can search for products using both text and image inputs for comprehensive product discovery.
• Dynamic Comparison Tables: The system generates real-time comparison tables based on user queries and product selections.
• FAQ-Enhanced Product Data: Each product includes comprehensive FAQ sections and enhanced data for informed decision-making.
• Session-Based Continuity: The platform maintains conversation context throughout user sessions for natural follow-up interactions.
• Mobile-Responsive Interface: Fully responsive design ensures an optimal experience across all devices and screen sizes.

2. Admin Users

• Admin Panel with Query Analytics: Comprehensive dashboard providing insights into user search patterns and query performance metrics.
• Product Management Workflows: Streamlined processes for adding, editing, and categorizing products with efficient workflow management.
• Category Attribute Management: Flexible system for managing product categories and their specific attributes for accurate classification.
• User Session Tracking: Monitor user engagement patterns, session duration, and interaction flows to optimize performance.
• Performance Metrics Dashboard: Real-time analytics showing system performance, recommendation accuracy, and user satisfaction metrics.

3. System Administrators

• Rate Limiting & Throttling Controls: Advanced controls for managing API usage and ensuring optimal system performance during peak periods.
• Security Monitoring: Comprehensive security monitoring tools, including threat detection and suspicious activity alerts.
• Database Optimization Tools: Built-in tools for monitoring database performance and optimizing vector search operations.
• API Documentation & Testing: Complete Swagger/OpenAPI documentation with integrated testing tools for development and integration.
• Containerized Deployment: Docker-based deployment system with CI/CD pipeline integration for reliable infrastructure management.
• Logging & Error Tracking: Comprehensive logging system with error tracking capabilities for quick issue identification and debugging.

Transform Product Discovery With AI-Powered Recommendation Systems

No matter the industry you represent, custom AI-powered recommendation systems are the ultimate solution that you must invest in. Why?

With custom AI recommendation engines, you can provide personalized product discovery experiences that eliminate brand bias and ensure variety in suggestions. This AI-powered system allows users to find relevant products through natural language conversations while preventing the repetitive results that plague traditional recommendation algorithms.

If you operate an eCommerce platform or curated marketplace, we recommend that you invest in custom AI-powered recommendation systems to enhance user engagement and product discovery. Such bespoke solutions enable you to create conversational shopping experiences with intelligent follow-up questions and contextual product comparisons.

Ultimately, it will help you provide diverse, relevant product recommendations that keep users engaged, no matter the size of your product catalog.

Get in touch with our experts today to strategize your artificial intelligence recommendation system.

• 
• 
•