中文 / EN
2022
2021
2020
2019
2018
2017
2016
2015
2008-2014
2022
2021
2020
2019
2018
2017
2016
2015
2008-2014

Mutual Assistance and Pushing the Limits

We donated a total of RMB 221.38 million to charities, and our employee "Pingfan Xiaozhu" personally donated RMB 138 million to support 23 charity projects of 15 charities to help the disadvantaged and promote social equity and development nationwide.

Fire-Flyer II has achieved the hardware and software architecture innovation of multi 800-port switches interconnection and core expansion subtree, breaking the physical limitation of Fire-Flyer Ⅱ Phase I and doubling the arithmetic capacity. The new hfai framework accelerates the model by 50-100%. The cluster runs continuously at full capacity with an average occupancy rate of 96% or more. Running 1.35 million tasks for the year, totaling 56.74 million GPU hours. 15.33 million GPU hours of idle computing power, or 27%, was used for research support.

Fire-Flyer II and Year of Public Welfare

High-Flyer AI spent one billion Yuan building Fire-Flyer II, an AIHPC with task-level time-sharing scheduling to share the heterogenous computing power. It has the state-of-the-art hardware: accelerator cards, 200Gbps High Bandwidth. Driven by our own software: the high-throughput parallel file system (3FS), the communication framework for distributed training (hfreduce), the collection of highly optimized DL primitives (hfai.nn), and the easy-to-use DL runtime (hfai), Fire-Flyer II is able to reach the limit of its hardware performance.

High-Flyer Quant became a director member of Asset Management Association of Ningbo.

Fire-Flyer I retired after 18 months' operation.

The High-Flyer Public Welfare Working Group was set up in order to contribute to the society in a professional and sustainable way.

Fire-Flyer I

The total investment of Fire-Flyer I was approximately 200 million Yuan. Equipped with 1,100 accelerator cards, it was put into use in the same year, which offered computing power support for AI research.

High-Flyer Capital Management (Hong Kong) Limited (High-Flyer Capital) was approved as QFII (Qualified Foreign Institutional Investor) with an aim to attract overseas funds to invest in the A-share market in the long run.

Ten-billion Quant

High-Flyer AI (Hangzhou High-Flyer AI Fundamental Research Co.,Ltd.) was registered, which was dedicated to the research on AI algorithms and its basic applications. The AI software and hardware team developed the High-Flyer Fire-Flyer I AI cluster, which was equipped with 500 GPUs and interconnected by 200Gbps High Bandwidth.

High-Flyer Capital Management (Hong Kong) Limited (High-Flyer Capital) was established and obtained the type 9 license in Hong Kong.

High-Flyer Quant became one of the ten-billion quants.

Won the Golden Bull Fund Award

The company decided to adopt the AI approach for development.

Computing power bottleneck occurs when single-machine training fails to meet the requirements of complex model computing. At the same time, the increasing training demand contradicts with the limited computing resources, thus a large-scale computing power solution is called for.

For the first time, High-Flyer won the Golden Bull Fund Award.

Overall AI Strategy

Keep expanding the AI algorithm research team and the AI software and hardware R&D team.

AI model calculation has been applied to almost all quantitative strategies by the end of 2017.

The First AI Model

The first stock position generated by the deep learning algorithm model was launched for real-time trading on October 21, 2016, using GPU for calculation. Before that, the algorithms mainly relied on linear models and traditional machine learning algorithms while the model calculation mainly relied on the CPU.

Beginning

High-Flyer Quant was established. It relies on mathematics and AI for quantitative investment. The high-spirited local team, without any overseas hedge fund background, is creative and diligent, and determined to become the world's top quant hedge fund.

Exploration

Founding team explores fully automated trading from scratch.