ReleaseBytes
gcp Google Cloud Blog ·

Malachyte details real-time AI architecture on Google Cloud for retail

blogaigcparchitectretailmediagcp-gkegcp-pubsubgcp-compute-engine
announcement

Malachyte, an AI-powered ecommerce recommendation platform, showcases how it leverages Google Cloud to solve the retail "cold-start" problem for new users. The platform uses neural networks with attention mechanisms, akin to large language models, to personalize product recommendations by continuously updating user vectors based on real-time interactions. Integrating Bigtable, Managed Service for Apache Kafka, Cloud Pub/Sub, GKE, and GCE enables sub-100ms inference times, helping retailers significantly boost sales. This case study demonstrates how even small teams can deploy sophisticated real-time continuous learning infrastructure using managed AI services on Google Cloud.

  • →Malachyte's AI approach to retail personalization
  • →Leveraging Google Cloud for real-time AI infrastructure
Notes (2) ›
  • Malachyte's AI approach to retail personalization

    Malachyte utilizes neural networks with attention mechanisms, similar to those in large language models, to personalize retail search and product pages. This approach captures the sequence of customer interactions to predict user intent and refine user profiles in real-time, effectively addressing the "cold-start" problem for new visitors.

  • Leveraging Google Cloud for real-time AI infrastructure

    Malachyte's real-time AI stack on Google Cloud streams behavioral events into Managed Service for Apache Kafka, with immediate updates to user profiles in Bigtable. Product catalog updates flow through Cloud Pub/Sub, while Malachyte agents and model inference run on Google Kubernetes Engine (GKE) and Google Compute Engine (GCE) to achieve sub-100ms recommendation cycles.

Read the original announcement →

https://cloud.google.com/blog/products/data-analytics/solving-retails-cold-start-problem-malachytes-recommendation-reinvention/

Related releases