Recently, a team of young researchers at CRRC, with an average age under 30, successfully developed Aizor - China's first industrial large language model (LLM) system that incorporates the scenarios across the entire chain of equipment manufacturing. The model significantly enhances the inference speed of LLM in various scenarios such as industrial design, manufacturing, and operations and maintenance.
To ensure seamless integration of the LLM into business applications, the team also independently developed the Huiyan AI platform. The researchers conducted in-depth field research on the production line to address inconsistent data standards and built a high-quality training dataset. Tackling the challenges of edge computing, they carried out 17 algorithm reconstructions and 42 rounds of solution iteration. Ultimately, they managed to address the key challenges such as complex data interaction and limited computing power, thus achieving a daily processing capacity of millions of industrial data entries and a millisecond-level response times.
Looking ahead, the team will continue to expand the application scenarios of the Aizor LLM and empower the high-quality development of high-end equipment manufacturing.
