Intelligent Applications

As described in the blog post Making Apps That Learn and Adapt the long-term vision for Spice.ai is to enable developers to easily build, deploy, and operate intelligent data and AI-driven applications.

With Spice.ai OSS, federated data and ML models are colocated with applications, creating lightweight, high-performance, AI-copilot sidecars.

The Intelligent Application Workflow

Dataset definitions and ML Models are packaged as Spicepods and can be published and distributed through the Spicerack.org Spicepod registry. Federated datasets are locally materialized, accelerated, and provided to colocated Models. Applications call high-performance, low-latency ML inference APIs for AI generation, insights, recommendations, and forecasts ultimately to make intelligent, AI-driven decisions. Contextual application and environmental data is ingested and replicated back to cloud-scale compute clusters where improved versions of Models are trained and fined-tuned. New versioned Models are automatically deployed to the runtime and are A/B tested and flighted by the application in realtime.