Actions:

Reformed the FE architecture and Introduce PostgreSQL vectorization storage to meet LLM consuming requirements.

Employed solutions such as RAG and multi-task to address model context length limitations.

Communicated with relevant teams within the company to discuss targeted optimization solutions and manage regulatory risks.

Achievements:

Promoted the standardization of key steps in large language model integration within the team, completed semantic optimization & vectorization of all component assets.

Design and developed the team's first AI-integrated feature, DataV Blueprint Copilot.

Implemented visual validation of suggestions and launched a monitoring solution for the pass rate of large language model generation, providing quantitative data of reasoning result quality.