In the AI era, data infrastructure is rapidly evolving from a “storage-centric” paradigm to a “compute-centric” architecture. With the explosive demand from large language models for vector retrieval, real-time stream processing, and large-scale analytics, databases are undergoing a new wave of technological innovation.
From April 21–23, 2026, YMatrix will participate in Postgres Conference 2026 in San Jose, Silicon Valley, engaging with global developers and architects to explore how to build a more efficient, stable, and intelligent data foundation for the AI era.
As a leading event in the global PostgreSQL ecosystem, this year’s conference is themed “Energizing People with Data and Creativity”, bringing together top developers and practitioners worldwide.
Unlike commercially driven conferences, Postgres Conference emphasizes deep technical exchange. The agenda features multiple core tracks, including Essentials (core internals), Dev & Ops, and Postgres Extensions Day, alongside sessions on Professional Development & Wellness, offering a holistic perspective from technology evolution to developer growth.
This event serves not only as a platform for technical exchange but also as a key window into the future direction of PostgreSQL.
At this conference, YMatrix will present two key technical sessions focused on high-performance execution engines and streaming architectures, showcasing innovations in data infrastructure for the AI era.
Session 1
YMatrix mxvector: How to Implement a High-Performance Pluggable Vectorized Executor in PostgreSQL
To address performance challenges in OLAP and HTAP workloads, YMatrix introduces mxvector, a high-performance pluggable vectorized execution engine.
Through lightweight kernel modifications and a plugin-based mechanism, mxvector enables a smooth transition from row-based to vectorized execution. Combined with batch processing, typed execution, SIMD acceleration, and optimized vectorized operators, it significantly improves large-scale analytical performance while maintaining compatibility with the native PostgreSQL ecosystem.
Session 2
PostgreSQL Streaming Model: From Implicit to Explicit – How Domino Unlocks Continuous Computing
Traditional Lambda and Kappa architectures often require complex data pipelines (Database + CDC + Kafka + Flink), leading to redundancy, consistency challenges, and high latency.
YMatrix introduces Domino, a native in-database streaming engine that enables unified batch and stream processing within PostgreSQL.
This session explores how PostgreSQL can evolve from implicit streaming to explicit streaming, redefining stream processing paradigms, simplifying real-time data architectures, accelerating AI application deployment, and achieving deep integration of storage, computation, and streaming.
The continuous evolution of database technology relies on global open-source collaboration. YMatrix aims to exchange insights with the global developer community while contributing its experience in hyper-converged architecture and kernel innovation.
In the AI era, YMatrix is committed to exploring the boundaries of data technology and advancing PostgreSQL into a versatile data foundation optimized for AI workloads.
