Recommendation Engine at Scale
12 weeks
Challenge
The e‑commerce team had a basic “customers also bought” setup. They wanted personalised recommendations on homepage, category, and cart—with sub-200ms latency at peak (Diwali-level traffic)—and the ability to A/B test models without engineering bottlenecks.
Solution
We built a recommendation service that combines collaborative filtering, item embeddings, and business rules. We served it from a low-latency API with caching and fallbacks so the site never blocks on recommendations. We plugged in an experimentation framework so product and data science can ship new models and measure impact without touching the core pipeline.
What we did
Results
“We went from generic blocks to personalised recommendations in three months. Latency held through Diwali and our data science team can now ship experiments without waiting on engineering.”
Other case studies
AI-Powered Transaction Platform
Leading Technology Company
Read case studyRetailCloud Migration & Optimization
Fortune 500 Retailer
Read case studyProfessional ServicesAI-Powered Document Intelligence
Top Professional Services Firm
Read case studyTechnologyPlatform Modernisation & API Layer
Mid-Size Technology Platform
Read case studyStart a project
Explore more work or get in touch.