NeuralCart

The Problem

A fast-growing e-commerce platform was struggling with a naive recommendation system that suggested random products based only on category.

The Solution

I designed a multi-signal recommendation engine combining collaborative filtering, semantic similarity via embeddings, real-time behavioral signals, and LangChain-powered contextual ranking.

Results

• +34% conversion rate on recommendation-driven purchases
• Sub-50ms p99 latency on 1M+ daily requests
• $2.1M additional ARR attributed to improved recommendations