Advisory Engineer, AI Application Development
Computer Technology - Artificial Intelligence Direction; Engaged in research related to artificial intelligence, data mining, and electronic document standardization in the computer open systems laboratory; During this period, I published a paper titled "Analysis of Traffic Congestion Based on Cluster and Correlation Analysis".
Applied for an invention patent "Processing Method, Device and Equipment of Application Data", Patent number: CN112559194A，Baidu
Python、PHP、GoLang
Machine learning, deep learning, CV, NLP
LLM SFT(LLama series,GPT,etc),Agents Application, SFT Corpus Construction,SD,Prompt Optimization、Agent Design、Autonomous Agent Design. A2A.
Distributed Architecture Design, Docker, Storage Design, MySQL, Redis, Kafka, OpenCV, etc
In the supply chain context, explore AI business implementation scenarios and conduct POCs for AI technical solutions tailored to business directions. Examples include DB-based data analysis, intent recognition for multi-data sources, knowledge base retrieval for data sources, SQL generation and data analysis hallucination resolution, Prompt Engineering, Prompt design, Agent design, Autonomous Agent design, and more.
Build AI teacher scene corpus, perform fine-tuning training on large language models, and create vertical large models for business scenarios. For example, AI teachers can teach English according to the teaching syllabus, with functions such as difficulty adaptation, theme gathering, expression correction, and content interpretation.
Explore AI Customer based on business, for saving labor costs, improving service efficiency, and improving service quality. It mainly solves three types of problem in daily work, including fixed answer response, answer of querying user data, and need to operate user data. Adopting Langchain's Agent architecture and implementing LLM based Agent, it includes Vector Library based Content Recall(Memory)、Planning Chain of Thoughts(Plan), and Function Call(Tools).
LLM basic services can encapsulate various large models, freely switch their protocols; Combining LLM with knowledge base to achieve precise content search, summarization, analysis, and other scenarios;
Build SD model basic services and AIGC middle platform services. SD model basic services mainly provide capabilities such as cultural and graphic graphics; The AIGC middle platform service mainly provides the ability to generate AI avatars, encapsulate specific styles of cultural and graphic graphics, and switch SD model basic services; Provides model capabilities in styles such as Disney and Anything v3, and supports related Lora models. Join ControlNet to achieve precise control of images, such as tiles_ Colorfix+Sharp preprocessor for color locking and sharpness adjustment of blocks; Canny preprocessor, edge detection model to achieve image contour detection; Etc;
Director of AI & Data Intelligence, Maintenance, Upgrade, and Iteration of Baidu APP's Collection Service with Billions-Level Data Volume, Including Architectural Refactoring, Design and Development of Baidu APP Bookshelf from scratch Architecture, Design and Development of C-end Counting Service 0-1 Architecture, Carrot Run Invoice, Push Service 0-1 Architecture Design and Development, etc;
Responsibility、Sense Of Honor、Research Ability、Innovation、Enthusiasm