担任中国联通网络AI中心资深算法架构师岗位，负责管理中心算法与大数据团队。工作内容包括以下三个方面，一是算法平台算法库建设，为通信业务场景提供基于时序预测、异常检测、图像识别、文本分析、通用分类预测等丰富的算法模型库及大数据特征数仓。二是建设基于知识图谱的通信网络拓扑，通过图算法模型构建智能排障、智能网络资源核配、ChatOPS等应用。三是针对通信领域网络移动网络、宽带等业务场景，基于人工智能技术进行科技攻关，在计算机视觉，时序异常检测、信号定位、流量分析等领域，形成基于智能巡检、自动化异常检测等算法能力，并在实际业务场景落地。 主要业绩如下： 一、形成基于深度学习DeepAR、Nbeats、Dlinear等复杂神经网络的时序预测算子库，通用通信网络KPI时序预测精度MAPE维持在20%以内，月调用量10亿以上；用户画像标签模型完成统一大模型建设，F1-Score在0.4以上，月度调用量2亿； 二、完成多省网络资源、告警信息知识图谱构建，智能故障定位召回率95%以上，Top3故障准确率80%以上。 三、智能巡检系统上线，具备OCR识别、传统仪表读数、设备检测等多项能力，平均ACC维持在90%以上。按照专业的简历介绍语言风格，润色一下，使用中文，并且纠正下语法错误及用词错误。

In the position of Senior Algorithm Architect of China Unicom Network AI Center, the individual is responsible for managing the center's algorithm and big data teams. Their responsibilities include the following three aspects:
1. Building algorithm platform and library to provide a wealth of algorithms such as time series prediction, anomaly detection, image recognition, text analysis, general classification prediction, and big data feature data warehouse to support the communication business scenarios.
2. Building a communication network topology based on knowledge graphs, using graph algorithm models to develop intelligent troubleshooting, intelligent network resource matching, and ChatOPS applications.
3. Conducting technology research in the field of communication network such as mobile network, broadband services, using artificial intelligence technology to form capabilities such as intelligent inspection, automated anomaly detection in the field of computer vision, time series anomaly detection, signal positioning, and traffic analysis, and landing them in actual business scenarios.

Achievements:
1. Formed a time series prediction operator library based on deep learning such as DeepAR, Nbeats, Dlinear, with a general communication network KPI time series prediction accuracy of MAPE within 20%, and monthly call volume of over one billion;
2. Completed the construction of a unified large model for user portrait tagging, with F1-Score above 0.4 and monthly call volume of 200 million;
3. Launched an intelligent inspection system that has OCR recognition, traditional instrument reading, and equipment detection capabilities, with an average ACC of 90%.