All work

Vector Search & Recommendations

Aura Smart Search API

An AI-powered search infrastructure providing semantic search and personalized product recommendations for large catalogs.

+35%
Search conversion lift
<15ms
Search latency
5M+
Indexed catalog products
Project details
Client
Aura Retail Group
Year
2025
Scope
Vector Search & Recommendations
Overview

Aura Retail Group aimed for a next-generation search API that understands semantic relationships, helping users find products despite typos or complex search queries.

The challenge

The challenge was keeping vector representations of millions of items synced in real time, while maintaining a sub-15ms response latency under heavy traffic.

Our approach

We vectorized product data using OpenAI Embedding models and indexed them on Pinecone. We built the API layer in Go, optimizing via gRPC protocol and Redis caching.

PineconeOpenAI EmbeddingsGoRedisgRPC
Results
  • Increased sales from search by 35%
  • Reduced search response latency to 12ms
  • Decreased zero-result search queries by 90%