Confluent Cloud for Apache Flink has introduced Flink Native Inference, Flink search, and built-in ML functions, offering a unified Flink SQL solution. These new features aim to simplify real-time AI development by eliminating workflow fragmentation, reducing costs, and enhancing security.
The new capabilities follow up on the GA release of the service last year. Adi Polak, a director at Confluent, elaborated on the pain points that Flink Native Inference, Flink search, and the built-in ML functions address for enterprise developers building AI applications:
Today, developers must use separate tools and languages to work with AI models and data processing pipelines, leading to complex and fragmented workflows. This results in a data pipeline sprawl, with data spread across many databases and systems. This lack of data integration and orchestration can lead to inefficiencies, increased networking and compute costs, and more development time, making it challenging to scale AI applications and see immediate value from AI investments.
The new features tackle these pain points by offering a unified, Flink SQL-based solution. Flink Native Inference allows any open-source AI model to run directly within Confluent Cloud. Polak explains:
These open-source and fine-tuned AI models are hosted inside Confluent Cloud, so there’s no need to download external files or manage the GPUs to run them. Instead of sending data to a remote model endpoint, Native Inference brings state-of-the-art open-source AI models to customers’ private data, allowing them to build event-driven AI systems using Flink SQL easily.
This approach eliminates network hops, enhances security by keeping data within the platform, and lowers latency for faster inference, which is crucial for real-time AI insights.
Flink search simplifies data enrichment across multiple vector databases, while the built-in ML functions make advanced data science capabilities accessible to a broader range of developers.
Regarding security, cost efficiency, and operational overhead, Polak emphasized the benefits of Confluent’s unified approach:
By unifying data processing and AI workloads, Confluent decreases organizations’ operational overhead by making it easier to develop and deploy AI applications faster. With Confluent Cloud’s fully-managed AI offering, companies can save costs with a hybrid orchestration of data processing workloads, more efficient use of CPUs and GPUs within a pre-defined compute budget, and benefit from having no cloud ingress or egress fees. There is also enhanced security since data never leaves Confluent Cloud, and proprietary data is not shared with third-party vendors or model providers.
According to the company, the new capabilities remove the complexity and fragmentation that have traditionally hindered real-time AI development – reducing the complexity of building real-time AI applications within enterprise environments.
