In the AI agent ecosystem, PyQ is a critical component for agents operating in high-stakes, low-latency domains such as automated finance or industrial monitoring. As agents move from simple chatbots to autonomous systems capable of real-time decision-making, they require the ability to query and analyze billion-row datasets instantaneously. PyQ allows these agents, which are almost exclusively written in Python, to treat the kdb+ database as a native extension of their own memory.
Kx is active at the intersection of the data and inference layers of the agent stack. By providing a bridge for LLM-based agents to perform complex vector searches and time-series analysis through PyQ, Kx enables the development of agents that can "remember" and act upon historical data at speeds that standard SQL or NoSQL databases cannot match. For builders using frameworks like LangChain or LlamaIndex, PyQ offers a path to integrate institutional-grade data infrastructure into their agentic workflows.
PyQ is a technical bridge that brings the Python programming language into the kdb+ database process. To understand why this matters, one must understand the environment in which Kx operates. The company’s flagship product, kdb+, is a high-performance time-series database and real-time analytics engine. It is the industry standard for high-frequency trading, where latency is measured in microseconds and data volumes are massive. Traditionally, interacting with kdb+ required knowledge of q, a terse and powerful vector programming language. While q is exceptionally fast for data manipulation, it has a steep learning curve and lacks the vast library ecosystem of Python.
PyQ solves this by allowing Python and q to share the same memory space. Unlike a traditional database driver that serializes data to move it between a database and an application, PyQ enables Python to access kdb+ data structures directly. This in-process execution means that a quantitative researcher can use NumPy, Pandas, or Scikit-learn on data stored in kdb+ without the performance penalty of data transfer. It effectively turns the database into a high-speed runtime for Python.
Kx was founded in 1993 by Arthur Whitney and Janet Lustgarten. Whitney, a legendary figure in APL and array programming, designed kdb+ to be compact and efficient. Based in New York with a global footprint, the company was acquired by FD Technologies (formerly First Derivatives) in stages, becoming a wholly-owned subsidiary in 2018. Over three decades, Kx has maintained a dominant position in the financial sector, counting most major investment banks and hedge funds as clients. PyQ emerged as part of the "Fusion" initiative, an effort to open up the Kx stack to modern developer workflows and languages.
In a market where many databases claim to be "real-time," Kx distinguishes itself through its obsession with hardware efficiency. The system is designed to leverage the underlying processor architecture, utilizing SIMD (Single Instruction, Multiple Data) instructions to process vectors of data at once. PyQ maintains this philosophy by mapping Python types to kdb+ types with minimal overhead. For example, a kdb+ list of integers is viewed by Python as a NumPy-like array. This architecture allows developers to build complex analytical pipelines that combine the ease of Python with the raw power of a vector-based database engine. It remains one of the few ways to perform complex machine learning on massive time-series datasets in real-time.
Direct in-process integration of Python and the kdb+ database.
Kx is hiring.