Case Study: Product Categorization
Agent for eCommerce Marketplace

A UK eCommerce marketplace using AI to automate product categorization and improve catalog accuracy. DjangoStars delivered eCommerce marketplace development services, building an intelligent agent and scalable architecture to streamline operations.
Let's discuss your project
legaltech
Team
6 team members
Type
eCommerce online marketplace
Industry
eCommerce
Platforms
Web
  • DjangoStars developed an AI-powered product categorization system featuring LLM-powered classification to automate item sorting in a large eCommerce marketplace.
  • The system enables marketplace product automation, automatically assigning new products to the correct categories, reducing manual review, and improving search accuracy.
  • The client needed to eliminate manual categorization, reduce errors, and enhance search and recommendation quality across the marketplace.
  • The main goal was to build a scalable, high-accuracy system capable of handling noisy input data, new product types, and large daily volumes with minimal human involvement.

Challenges

Challenges
  • Inconsistent and poorly structured product data from multiple vendors
  • High misclassification rate (over 20%) impacting search relevance
  • Cold-start problem for new or rare product categories
  • Strict latency requirements for real-time product publishing
  • Need to scale reliably to tens of thousands of new listings per day

Technologies we use

Technologies we use logo
Technologies we use logo
Technologies we use logo
Technologies we use logo

We Delivered

01 Designed a hybrid AI categorization pipeline combining rules, embeddings, and LLM reasoning.

02 Built a robust data preprocessing and normalization layer.

03 Implemented active learning tools for model improvement.

04 Optimized inference performance with caching and serverless deployment.

05 Delivered a scalable solution integrated into the client’s platform.

06 Leveraged AI in eCommerce marketplace development to automate large-scale catalog operations.

07 Improved eCommerce product categorization accuracy to reduce manual review and enhance search relevance.

Results of the Collaboration

Classification accuracy improved from ~71% to ~92%.

Manual review workload decreased by ~75%

Click-through rate (CTR) on search results increased by ~18%.

System now handles >50,000 new listings per day efficiently with minimal human intervention.

DjangoStars builds intelligent AI solutions for eCommerce marketplace platforms, helping businesses eliminate manual catalog processing, reduce classification errors, and scale product operations efficiently. We deliver LLM-powered systems that improve search accuracy, automate workflows, and support sustainable marketplace growth.

Testimonials

They have a ton of experience in the earlier stages of taking design concepts from a founder’s head and making it a reality. They’ve done the database design, the backend development and design, and the frontend development and design. The application has launched.

They’ve been in business for over 10 years, so they’re stable. They’ve been steadily growing since I met them. I’ve been to their offices and seen the culture of the company. They have a really good culture

5.0
More eCommerce cases More UK cases