Ecommerce: Helping Shopify Switch from Elasticsearch to an AI-Native Vector Search Engine

Search is the backbone of modern commerce. And nobody knows that better than Shopify. At Shopify’s scale, search is not a mere feature. It’s infrastructure. 

Millions of queries flow through the system every hour, especially during peak events and the holiday season. When shoppers search, they’re not typing keywords. They’re expressing intent.

But, Shopify’s legacy stack was nearing its limits, especially in an era where discovery was becoming semantic, vector-based, and AI-native. 

The decision to evolve was a no-brainer, but the real challenge was migrating the core discovery engine of a global commerce platform, and all that without disrupting performance, quality, or peak-season traffic.

Here is how SmartCat helped them pull it off without a hitch.

Problem

Shopify’s legacy Elasticsearch infrastructure had become a bottleneck. While it was still reliable for traditional keyword search, it lacked native vector support and the semantic capabilities required for modern, intent-driven discovery.

This was especially important, because shoppers nowadays are used to having conversations instead of typing in keywords or filtering.

As search expectations evolved, the architecture began to strain under:

• Increasing throughput demands
• Tight latency requirements at global scale
• More complex, intent-driven discovery use cases

To combat this, Shopify had acquired a superior AI-native search engine purpose-built for vector search and semantic retrieval from SmartCat. But acquisition alone doesn’t operationalize a platform.

Internal teams faced a significant knowledge gap around the new engine’s architecture and performance characteristics. 

Migrating core discovery infrastructure required deep familiarity with both the legacy system and the AI-native platform, and the migration had to be executed without downtime, regression, or missed SLOs.

Solution

Our engineers, who originally built the newly acquired search engine, joined the effort to provide the technical leadership needed for a high-stakes migration.

We led the architectural transition from Elasticsearch to the AI-native vector platform, redesigning critical components to meet Shopify’s high-throughput, low-latency requirements. This included:

• Refining inter-service communication protocols
• Implementing sharding strategies tailored for vector workloads
• Optimizing indexing and query execution paths for sustained performance

Beyond infrastructure changes, we focused heavily on validation. To guarantee a seamless transition, we developed a comprehensive custom testing framework that enabled:

• Real-time validation of semantic search quality
• Continuous measurement of indexing speed and query latency
• Stress testing under extreme traffic conditions

With this framework, we made sure that performance gains never came at the expense of relevance or system stability.

Results

The migration was completed on schedule, successfully transitioning Shopify’s core discovery infrastructure to the AI-native vector engine.

Thanks to architectural optimizations, the platform consistently exceeded Shopify’s stringent Service Level Objectives across:

• Query throughput (QPS)
• Indexing speed
• Query latency
  

Most importantly, the system was fully operational and optimized ahead of the peak holiday season. During Black Friday and Cyber Monday, the new platform handled massive traffic volumes without a single critical disruption, maintaining high availability and low latency throughout.

The result is a stable, high-performance foundation for AI-native commerce. Shopify’s internal teams now operate and scale the platform independently, backed by a testing framework that safeguards both speed and quality.

About the Client

Shopify is a global commerce leader providing essential internet infrastructure for millions of merchants across 175 countries. The platform powers businesses ranging from independent entrepreneurs to major enterprise brands.

As part of a broader strategic shift toward AI-native commerce, Shopify is replacing legacy infrastructure with proprietary, AI-driven technologies.

This includes advanced discovery engines and vector-based architectures designed to deliver hyper-personalized shopping experiences.

Technologies Used

The migration relied on a distributed, production-grade stack capable of supporting high-performance vector search at a global scale:

• Languages: C++, Java, Python
• Cloud & Orchestration: Google Cloud Platform (GCP), Kubernetes
• Streaming & Data Pipelines: Kafka

If your Ecommerce platform still relies on keyword search alone, you’re likely missing opportunities to connect shoppers with the products they actually want.

Discover how our SmartCat’s marketplace solutions can help your ecommerce platform improve product discovery and search performance.

Also, you can book a consultation or sign up for one of our workshops, and together, we will help you determine what sort of solution is would work best for your enterprise.