Zhongke Jiahe Technology

Zhongke Jiahe Technology, established in July 2023, is an AI ecosystem software infrastructure provider that focuses on compiler technology to bridge the gap between domestic computing power chips and the implementation of large models. The company aims to deliver generalized, low-cost, and high-performance computing power resources for China's AI large-model industry.

The company was founded by Cui Huimin, who graduated from the Computer Science Department of Tsinghua University and leads the compiler team at the Institute of Computing Technology, Chinese Academy of Sciences (CAS). The core team originates from this CAS compiler group and possesses over 20 years of development experience, having contributed to the compilers for several domestic chips, including Loongson, Sunway, Cambricon, and Huawei Ascend.

Zhongke Jiahe has developed a product matrix that includes a heterogeneous native large-model inference engine (SigInfer), a heterogeneous native fine-tuning engine (SigFT), and an operator automatic generation and translation tool (SigTrans). These tools are designed to be compatible with multiple domestic AI chips, providing a unified, high-performance interface that masks hardware differences. The main offering, the large model inference optimization tool, enhances the performance of domestic computing power through compiler optimizations. Another key product is a CUDA-compatible compiler that addresses compatibility challenges for domestic chips, enabling the automatic translation and optimization of operators. This allows applications from platforms like NVIDIA's to be seamlessly ported to domestic chips.

The firm's business model is centered on providing these foundational software tools to a range of clients, including domestic AI chip manufacturers, hardware equipment vendors, server manufacturers, internet companies, and intelligent computing centers. By creating a unified software platform that supports various CPUs, GPUs, and NPUs, Zhongke Jiahe helps clients reduce development costs and accelerate time-to-market. Since its inception, the company has secured significant funding, including a seed round in August 2023 and a Pre-A1 round in January 2025, attracting several tens of millions of yuan and numerous commercial orders.

Keywords: compiler technology, AI software infrastructure, heterogeneous computing, large model optimization, AI chip ecosystem, CUDA compatibility, compiler toolchain, AI inference engine, fine-tuning engine, operator translation, high-performance computing, domestic AI chips, neural processing unit, GPU optimization, computing power solutions, AI foundational software, cross-platform software, system optimization software