Retail Giant Leverages GenAI
for Intelligent Product Discovery
Customer Background
A global retail and lifestyle eCommerce company with over 50 million monthly users across North America and Europe. The brand offers over 300,000 SKUs in fashion, electronics, home, and beauty categories.
Challenges
The customer faced declining conversion rates and high drop-offs during product searches. Their legacy keyword-based search system couldn’t understand intent or adapt to evolving user behavior. Static recommendations lacked personalization and failed to surface relevant products, especially in long tail or multi-intent queries. Despite a large product catalog, customers struggled to find what they needed. Their engineering team also faced issues integrating AI/ML due to legacy infrastructure and performance bottlenecks on mobile.
Solutions
Exilon implemented a GenAI-powered conversational assistant using a fine-tuned OpenAI GPT model, integrated with the product catalog and search APIs. We used FAISS for semantic search and product embeddings trained with sentence-transformers. Behavioral data pipelines were built in Databricks, using historical clickstream and purchase data to fine-tune product matching. The assistant was deployed as a widget on mobile/web with fallback logic to traditional search. Performance was optimized with caching layers and fast inference APIs, reducing latency by 45%. Our solution plugged into their CI/CD workflow and ran seamlessly across regions.
Benefits
- 38% increase in product discovery rate
- 27% boost in conversion on mobile/web platforms
- 22% drop in cart abandonment
- 45% reduction in search latency