Case Study: The Ikigai Lab achieves scalable interactive AI workflows with Anyscale and Ray Serve

How Ikigai Labs Serves Interactive AI Workflows at Scale using Ray Serve

The Ikigai Lab built an AI-augmented spreadsheet and analytics platform that lets users create interactive, mission-critical data pipelines with custom Python code. Their challenge was to make these workflows both highly scalable and instantly browsable, while still supporting many user-defined scripts with conflicting dependencies and a collaborative, spreadsheet-like experience. Anyscale’s Ray ecosystem, especially Ray Core and Ray Serve, was the foundation for addressing these needs.

Anyscale helped The Ikigai Lab serve arbitrary Python code at scale with Ray Serve, manage per-deployment Conda environments, and expose HTTP endpoints for low-latency “peek” operations without Ray task-submission overhead. The team also used version-aware deployments to avoid race conditions during concurrent updates and scaled popular services with replica-count controls. As a result, Ikigai achieved sub-second browsing of intermediate datasets, smoother collaborative deployments, and a more reliable, scalable interactive AI workflow platform.

The Ikigai Lab