Confluent Enhances Apache Flink Capabilities for Streamlined AI Development

AI Development on Flink Gets More Accessible with New Features

Confluent, known for its real-time data streaming solutions, has expanded the capabilities of Confluent Cloud for Apache Flink with new tools designed to ease the creation of AI-powered applications. The newly launched features—Flink Native Inference, Flink search, and built-in machine learning functions—are aimed at reducing the friction developers face when building and deploying real-time AI systems. These tools offer a more consolidated and less fragmented environment for managing models and streaming data.

With Flink Native Inference, teams can now run open source or fine-tuned AI models directly within Confluent Cloud, eliminating the need for separate infrastructure. The data remains within the platform, offering a more secure and cost-efficient solution. This capability targets a common challenge where developers often rely on multiple tools and environments to execute inference workflows, leading to delays and operational complexity.

Unified Vector Search and Streamlined ML Integration

Flink search simplifies the process of accessing and querying real-time data across various vector databases like MongoDB, Elasticsearch, and Pinecone. Rather than performing manual data consolidation or building complex ETL processes, developers can use a single interface to enable large language models (LLMs) to retrieve contextual information. This method supports more accurate responses and reduces the risk of hallucinated outputs.

The introduction of built-in ML functions within Flink SQL further expands the accessibility of AI development. Tasks like anomaly detection, forecasting, and data visualization can now be performed without requiring advanced data science skills. These additions open up real-time analytics to a wider audience of developers, encouraging quicker decision-making and more responsive applications.

Enterprise Use Cases and Market Adoption

According to Steffen Hoellinger, Co-founder and CEO of Airy, the platform helped them integrate real-time organizational knowledge into AI workflows more effectively. He noted that by using Flink AI Model Inference along with vector databases, they were able to support RAG-based systems and schema intelligence more efficiently, resulting in improved productivity and operations.

These enhancements are part of Confluent's broader effort to unify batch and real-time data processing under a single, fully managed, serverless stream processing platform. By reducing the need for multiple systems, the platform enables organizations to simplify their technology stacks and focus on delivering meaningful AI use cases.

Early Access and Additional Updates

The new features are currently available through an early access program for Confluent Cloud customers. Alongside the AI and ML tools, Confluent also announced updates such as Tableflow, Freight Clusters, a new Visual Studio Code extension, and the Oracle XStream CDC Source Connector, further expanding the platform’s integration capabilities.

Stewart Bond, Vice President at IDC, highlighted the importance of integrating reliable and contextual real-time data into AI systems. He emphasized that platforms like Flink, offering unified inference and search functionalities in a managed cloud-native environment, could significantly impact the practical application of generative and agentic AI in enterprises.