Tong Shen
ByteDance AI Development Platform R&D Engineer
He is a member of the framework group of ByteDance AI development platform, graduated from Peking University, focusing on the design and implementation of large model application development frameworks, and is one of the core developers of the Eino framework (https://github.com/cloudwego/eino).
Topic
From Components to Orchestration - Design Practices for Large Model Application Development Framework
This talk discusses the connotation of the field of large model application development, analyzes the key problems that need to be solved by the development frameworks in this field, introduces the solutions in detail by taking Eino as an example, and demonstrates the problem solving process and methodology through the specific process of building intelligences, and sums up the experience in the process of designing and developing the frameworks. It is hoped that the audience can gain the ideas of solving some important problems in the field of large model application development, as well as obtaining experiences and thoughts on framework design through the design choice of Eino framework. Outline: I. Domain of Large Model Application Development 1. Domain Modeling 2、Generic expression of business model 3、Generic flow of data Second, the big model application development framework to address the key issues and solutions 1、Component abstraction and realization, Eino's approach 2, business process orchestration capabilities, Eino's approach 3、Data flow processing, Eino's approach Case: Eino assistant intelligent body 1、Functional description and components used 2、Business process and organization 3、Stream processing effect Design experience 1、Practice-driven 2、Cautious abstraction 3、Kernel + extension