Kryera2

Kryera2 is a cross-platform modular software framework designed for building data processing pipelines and machine learning workloads. It is the successor to the Kryera project and is developed by the Kryera Foundation, with an emphasis on performance, portability, and extensibility.

The architecture centers on KryeraCore, the execution engine, and KryeraDSL, a domain-specific language that compiles pipelines into a data-flow graph. KryeraPlugins provides pluggable algorithm implementations. The framework supports streaming and batch data, with labeled tensors and records as the data model. It also emphasizes a clean separation between pipeline definition and execution, enabling flexible experimentation and deployment.

Performance features include multi-threaded scheduling, SIMD optimizations, and optional GPU offload via CUDA and OpenCL. It offers bindings for Python, C++, and Rust, facilitating integration with existing data stacks. The project supports common data formats such as JSON and Parquet, and provides a lightweight runtime for containerized environments.

Kryera2 was released in 2022 as an evolution of the original Kryera, with a focus on richer extensibility and improved resource management. It is distributed under a permissive open-source license and has an active community contributing plugins, tutorials, and example pipelines.

In practice, Kryera2 is used for ETL pipelines, real-time analytics, and serving machine learning models in production. Its ecosystem favors teams that value portability across platforms and the ability to integrate custom algorithms without reworking pipelines.