Featureform is joining Redis, adding a robust framework for managing, defining, and orchestrating structured data signals to Redis’ speed and simplicity. Featureform will be integrated into Redis’s feature store offerings within the real-time data platform for AI, alongside the fastest benchmarked vector database powered by Redis Query Engine and the advanced semantic caching service, Redis LangCache.
Featureform is a virtual feature store built on Apache Iceberg, with prebuilt integrations for Snowflake, Clickhouse, Spark, and more. It addresses a major challenge in production AI: delivering structured data to models quickly, reliably, and with full observability.
Simba Khadder, Featureform’s founder and CEO, said: "Featureform acts as a ‘motherboard’ for AI infrastructure. It connects data sources, compute engines, and online stores into one cohesive system. Featureform integrates with the tools you already use, including Spark, Snowflake, Iceberg, Kafka, and of course, Redis."
The combined Redis + Featureform solution delivers an end-to-end system for managing and serving structured data in AI applications. For teams already using Redis as an online store, Featureform adds a flexible, declarative system for defining and orchestrating features across training and inference workflows.
Redis + Featureform enables teams to:
- Define features once and use them across environments
- Orchestrate batch, streaming, and real-time workflows in one framework
- Serve features instantly via Redis as the online store
This integration supports GenAI, ML, and agentic systems with a fully managed, Redis-native feature pipeline from raw data to low-latency inference. Unlike traditional feature stores, Featureform provides orchestration without storage lock-in, integrating with existing tools while centralizing feature management.
Featureform also builds on community efforts such as EnrichMCP, exploring agent-driven retrieval of structured data on demand. Redis + Featureform provides the simplest path to building end-to-end structured data pipelines while continuing to support other feature stores, including Feast and Tecton. Redis remains open, flexible, and compatible with any feature store, offering Featureform as an optional Redis-native solution.
The industry is moving beyond naive retrieval-augmented generation (RAG) and basic LLM orchestration. Teams now need to enrich AI prompts with user profiles, financial metrics, risk scores, and business logic. Featureform treats all inputs as features — including tabular values, embeddings, prompts, and model inputs — providing a unified language for defining and managing AI system building blocks.
Redis + Featureform bridges the gap between fast vector search and rich structured context, delivering the right information at the right time. Both embeddings and structured features can be served from the same platform, with the speed, reliability, and simplicity expected from Redis.