BACK TO PROJECTS

BACK TO PROJECTS

AI STRATEGY

E-COMMERCE

Product Classification System

Setup and use of LLMs on secure servers for mapping legacy product data in Arabic to English UNSPSC codes. 

CLASS ACCURACY

99.8%

DELIVERY TIMELINE

8 Weeks

REGION

Saudi Arabia

CLIENT PROFILE

Industry

E-commerce Strategy

Location

Saudi Arabia

Year

2024

TECHNOLOGY STACK

  • Llama 3 3.2
  • Python

The Challenge

A significant challenge was managing a large volume of unstructured product data in Arabic, which urgently needed standardization to UNSPSC codes for consistent reporting while following the strict data transfer policy.

"Unstructured Arabic product data needing UNSPSC standardization while no external APIs could be used."

The Solution

We setup new secure server using high end NVIDIA GPUs with Llama and implemented an advanced AI-driven product classification system leveraging Llama 3 3.2 for sophisticated multi lingual text processing.

Text Embeddings

Converting translated Arabic descriptions into machine-readable embeddings.

Cosine Similarity

Matching embeddings with UNSPSC code database for automated classification.

Llama 3 3.2 Model

Sophisticated text processing of Arabic product descriptions.

Full Impact Analysis

1

Successfully integrated legacy product data in local Arabic dialect with English UNSPSC code based current systems, breaking down data silos.

2

Enabled continuous reporting across historical and current data.

3

Improved data standardization and accessibility, streamlining operations.

FUTURE ROADMAP

Further extention of mapping system over other languages for other geographies.

Ready to build your success story?

Initiate Consultation

Partner. Build. Scale

Partner. Build. Scale

READY TO START ?

Let's discuss how we can engineer it into reality. We are ready to partner with you.

All rights reserved. WBOS.
POWERED BYOPMIZ