AI/ML Functions
Unleash intelligent real-time AI by unifying AI/ML with data processing in Flink. Perform model inference, vector embedding, anomaly detection, forecasting, and more directly on data streams.
Accelerate time to ROI
Use your model of choice
Zero AI/ML expertise required
What Are AI/ML Functions?
AI/ML Functions are fully managed, built-in capabilities in Confluent Intelligence that let you run AI and machine learning tasks directly on streaming data with Flink SQL. From model inference and generating embeddings to anomaly detection, they help teams turn live data into real-time predictions, decisions, and action – without stitching together separate data and AI infrastructure.
✔ Real-time model inference using LLMs of your choice
✔ Continuously updated vector embeddings for your vector databases
✔ Easy ML for instant insights
✔ Ideal for building first AI use case and triggering agentic workflows
Join AI Day in Toronto
MAY 26 – Learn, build, and hack at AI Day Toronto! Master context engineering, MCP, Flink and more to bring your AI agents and apps to life.
Real-Time AI/ML Use Cases
AI-powered data enrichment
Enrich data with AI-generated insights and summaries, and context in real time.
Retrieval-augmented generation (RAG)
Generate vector embeddings for RAG to ground AI responses so outputs are relevant, accurate, and free of hallucinations.
Event-driven Streaming Agents
Build agents that observe live business signals, call LLMs to reason in real time, and take action as events happen.
GenAI applications
Power real-time GenAI applications with the streaming data and inference they need to deliver value.
Anomaly and fraud detection
Detect unusual patterns and suspicious activity in real time so teams can respond faster and reduce risk.
Sentiment analysis
Understand customer sentiment from live text streams to improve service, engagement, and decision-making.
Forecasting
Predict demand, trends, and outcomes like continuously using live data instead of stale batch snapshots.
PII detection and redaction
Detect PII in data streams and flag or redact it in real time to protect privacy and ensure compliance.
Get started and launch ML in minutes by using the step-by-step Quickstart.
Remote AI Model Inference
Connect to external AI/ML models – from OpenAI and AWS Bedrock to Vertex AI, Azure AI, Anthropic, and Fireworks AI – and run inference directly in Flink SQL on live data streams. This lets you add real-time predictions, reasoning, and RAG to your applications without stitching together separate AI and data processing infrastructure. View docs >
Built-in AI/ML Functions
Turn AI/ML use cases and complex data science tasks into simple Flink SQL functions for anomaly detection (ARIM), forecasting (TimesFM), sentiment analysis, PII detection and redaction. Teams can move from raw data to real-time intelligence faster, with no ML model building or deep data science expertise required. View docs >
Preprocessing Functions
Prepare streaming data for better downstream results with built-in preprocessing functions such as bucketizing, normalization, and text splitting. By transforming features into model-ready representations directly in Flink, you can improve data quality and reduce extra preprocessing work outside the pipeline. View docs >
Ready to build real-time AI? Sign up and receive $400 to spend during your first 30 days.
Use your model of choice, apply built-in functions, and move from raw data to production use cases faster.
- ✔ Use external, hosted, or custom models for real-time inference on streaming data
- ✔ Turn complex AI/ML tasks into simple Flink SQL functions for faster time to value
- ✔ Reduce complexity and infrastructure overhead with a unified platform for data and AI workflows