缺

founder and CTO of einplus, has been programming for more than 30 years, 12 years of experience in the Internet industry, and currently in the commercial dialogue robot entrepreneurship. Former Chief Architect of Douban and Chief Architect of CreditEase Big Data Innovation Center. A veteran Python developer who is one of the founders of the Chinese Python User Group (CPUG). He is the first full-time employee of Douban, and responsible for the system architecture. He has experienced the process of watercress from 100,000 to hundreds of millions of PVs, and has accumulated a profound experience in large-scale website architecture. He has led the construction of Douban internal PaaS system DAE and Credit Suisse financial container cloud Lain, with rich experience in container cloud and microservice architecture. His current focus is on chatbot, cloud computing, container technology, microservices, and security technologies.

Python is an open-source language where everyone can contribute.  Over the years the workflow used by both contributors and core developers changed dramatically, from patches sent via email to pull requests sent on GitHub.  With this talk I want to unveil what currently happens ''behind the scenes'' of CPython and how you can get involved and be part of the open source community that allows Python to be one of the most popular programming languages.

Ezio Melotti is a software engineer with almost 20 years of experience in software development and programming with several different languages. Born in Milan (Italy), he decided to move to Turku (Finland) in the year 2007, where he graduated and then taught at the Turku University of Applied Sciences. In 2009 he became CPython Core Developer and contributed over 1000 changesets over the years. He is also the maintainer of the bug tracker and the html package and did extensive work on the test suite, on the documentation, and on the standard library.

1. Python's ecology in the field of software testing
2. The birth background of HttpRunner
3. The development and design ideas of HttpRunner
4. Automated test program based on HttpRunner
5. Successful practice case based on HttpRunner

Leo Lee (debugtalk), DebugTalk Blogger, the author of HttpRunner. 7 years‘ experience in software testing industry, heavy Python enthusiasts, always committed to Python in software testing field of application and promotion. Embracing the Zen of Python, pursuiting of simplicity and elegance in code, as well as in life. Currently in DJI is responsible for testing and development work.

This speech will share the most frequently encountered problems in CPython development and the tools for debugging problems in your own work scenario; also, it will talk the thinking about Python debugging and monitoring in future.

updating…

Extend the use of cpython vm through metaprogramming to improve code reuse rate, abstract ability and runtime program execution efficiency. Also to introduce How to introduce JIT and type checking tools in a compatible cpython ecosystem.

From MSRA, has very deep research on features and underlying of Python language

This speech will systematically introduce the causes of GIL and the side effects it brings. And bring a complete set of solutions to properly circumvent GIL to better improve the performance of Python online.

He started learning Python in 2005 and have shared millions of words of Python notes and translated documents. Former Tudou Assistant Architect, former Crust Network Founding Technical Director, former deepglint engineer, former RoboTerra Software Engineering Director. Fans of Radio/model/robot.

This topic will share the current status and ideas for useing Rust code in Python.

Python Backend Developer at day, Rustacean at night. Author of PyO3, active in the Python community and the Rust community, also enthusiastic and hardworking in open source enthusiast.

updating

Joined Zhihu in 2013, he was the head of the home feed stream engineering technology. During the project of the feed stream, the response time P95 was reduced from 1.6s to 700ms through the optimization of the framework. The stability was improved from 99.9% to 99.99% through the development specification. Responsible for the Community Architecture Group in 2018, leading the team to rewrite the main business modules with Golang, saving machine resources 75%

Python is a programming language with multiple inheritance capabilities. In object-oriented inheritance, the invocation of methods is closely related to the order of inheritance. Python uses the C3 linearization algorithm to calculate the MRO as the only criterion for the method call sequence, and understanding MRO can help us in Python object-oriented programming.

updating

This presentation will introduce the knowledge and skills of using Python for web crawler development in five aspects: crawling, parsing, storing, anti-climbing, and acceleration. It introduces how to take different measures to efficiently capture data in different scenarios, including Web crawling, App crawling, data storage, agent purchasing, verification code cracking, distributed crawling and management, intelligent parsing, etc., and also introduce some commonly used toolkits in combination with different scenarios. All of the speech content are the summary of the experience of the speaker since the web crawler research process.

Qingcai Cui, master of Beihang University, author of 《Python3网络爬虫开发实战》, Blog bloggers of cuiqingcai.com, the number of blog reading about Crawler has exceeded one million, Big Data Engineer of Microsoft China, lecturer of Tianshan Intelligent and Netease Cloud Classroom, and currently engaged in the research of conversational chat direction.

1.Basic knowledge and trends of artificial intelligence
2.Picture classification and fine-tuning practical case
3.Face recognition direction practical case
4.Numerical analysis direction practical case

Jinhong Lee, main engineer of Code doctor studio, founder of the big snake intelligence, chief artificial intelligence scientist of Beijing Pai Network Software, has released two video tutorials on the CNTK deep learning framework, has published two best-selling books, 《深度学习之Tensorflow：入门、进阶与实战》and《python带我起飞：入门、进阶、商业实战》.

Mars - Matrix-based unified computing framework, the goal to bring the ecology of PyData to a distributed execution environment, currently supports the Numpy interface, can perform multidimensional array operations in a distributed environment, it supports running on the GPU, and supports two-dimensional sparse matrices. Currently, Mars is compatible with 70% of the common Numpy interfaces. This sharing session introduces our current work, introduces Mars'' implementation principles, and looks ahead to future superstructures including DataFrame and machine learning.

His main job is to combine the Python ecosystem with the big data environment. The past work involved writing the pyodps DataFrame framework, which allows users to write code similar to pandas, but supports execution on MaxCompute bi data platforms, pandas, and traditional databases. His current work is mainly Mars, which is a matrix-based unified computing framework. Its goal is to bring the PyData ecosystem to a distributed execution environment. Currently, it supports the Numpy interface and can perform multidimensional array operations in a distributed environment. It supports running on the GPU. And supports two-dimensional sparse matrices. Pandas and scikit-learn will be distributed in the future.

1.Introduction of TPU;
2.A tutorial of training/inference a DL/ML model on Google TPUv2 2.1 example 1, 2.2 example 2;
3.Difference between TPUv2 and GPU and trouble shooting of TPUv2. 4 Benchmark of TPUv2 and GPU

Python/Go/C++ developer, responsible for the landing and best practices of cloud computing, big data and deep learning technologies in large enterprises. He has worked for artificial intelligence unicorn Face++ and AWS, Oracle and other companies. The main architecture and developer of the face recognition system for the coffee-free unmanned convenience store, AWS Big Data Expert EMR Subject Matter Expert.

The new asset types (digital currency and vouchers) brought about by the blockchain technology revolution and its decentralized infrastructure are bringing dramatic changes to the global asset management industry. With its mature large-scale financial application background and active and active user community, the Python language has become the preferred solution in the blockchain asset management business. This presentation aims to share the application of Python in the various ecological aspects of blockchain asset management and to explore future directions.

Xiaoyou Chen, founder of Shanghai Wayne Kailuan Software Technology Co., Ltd., has worked as a quantitative trader and fund manager in large-scale quantitative private equity funds and developed the vn.py quantitative trading framework, which is currently one of the world''s largest open source quantitative trading projects (Github Star 6071). His current focus is mainly on development of quantitative trading systems in the blockchain and digital currency sectors.

1. The overall evolution and architecture of the data science platform from 0 to 1
2. 0.25 era - "Python + Hive + Shell" mixed programming to solve the cumbersome human ETL
3. 0.5 era - "Django+ElasticSearch" increased customer search and BI analysis support
4. 0.75 era - "Django+PyHive+PySpark" increases the operation and machine learning task flow of big data clusters
5. 1.0 era - Scheduler scheduling integration and platform, cluster micro-service
6. Various external technical resistances and pitfalls encountered in the evolution path from 0 to 1
7. Summary and outlook'

One of the executors of the AI Empowerment Program, working as a guest lecturer in cloud computing and big data related fields in several universities in Shanghai, first contact with Python in 10 years, and fully embraced Python and deepened the data science community in 15 years.

In recent years, PWA and miniprogram become more and more popular in app developments; in the mean while BaaS and FaaS also be loved by more and more developers. In this session, the speaker will introduce how ifanr team build a serverless platform which designed for miniprogram developers, using Python and Django.

Start-up & Architect @ Guangzhou, China CTO @ ifanr.com, in charge of cloud.minapp.com, a serverless platform designed for miniprogram apps.Art lover / addicted in creative activities.

With the increasing number of Internet attacks, network information security has gradually attracted people's attention. Python has become the most popular programming language for hackers. Whether it is network attack or defense or software cracking, Python is very simple to help security personnel accomplish many things, such as: automation Attacks, etc. Through this topic, you can learn about Python's things in the information security field.

Taoxun Kong / K0r4dji, CSO of Beijing Ding Niu Technology Co., Ltd., founder of Breaking Security Team, which is a famous information security team in domestic, book expert consultant of People Post Press-Asynchronous Community-Information security field.

With the formation of ecosystems such as containerization and cloud native, the value of CI/CD in the software development process is becoming more and more prominent. More and more companies need to transform and optimize this, but thousands of people have thousands of faces, transformations and practices often have many problems, both in the personnel and in the tool chain. This time is mainly to share the best practices we have summarized in production, and to analyze the problems encountered in this process, and to share in detail the technical details, as well as the optimization, problem analysis, and architecture in the process and the reflections and conclusions we have made, I hope that everyone can inspire them and promote their respective development in the production environment.

Jintao Zhang, senior devops developer of NetEase, responsible for construction of containerization and automated operation and maintenance system.

The development of Web technology has been quite complicated so far. The related projects are “small and no internal, and they are nothing short”. Various front and rear end frames and middlewares are numerous and numerous. Many introductory developers have even worked for three or five years, limited to the use of a certain two frameworks, and regarded their own fancy framework as a standard. They are only skilled in the interface after the framework is packaged. Why does the framework exist and why in that design? When encountering difficult problems, and how to solve and optimize, it is not enough. This topic is based on the history of Web technology development, helping listeners to clearly understand the role and key points of Web-related core technologies, so that they can have clear goals and clear ideas when solving various problems encountered in learning work in the future.

He has been working for more than 5 years, focusing on the Python technology stack server architecture design and framework development. He used to be the R&D engineer of the Lea cloud computing PaaS platform and the head of the data department of Cheetah Mobile AILab. He is currently the director of engineering of the metasota. In the spare time, he often exchange ideas and guide new people in the technical community with Python developers.

As a popular Python web framework, many developers like Flask''s simplicity and flexibility, and often compare it to bulky Django. Based on these characteristics, we can say that Flask is free, and free Flask will make your web development more free. Specifically, the free here is manifested in many aspects, including the extension of program functions, the definition of routes, the organization of project structures, and the design of program patterns.

The author of Python Web Development with Flask.

This speech focuses on the work experience of speecher and illustrates the practical use of Python in the embedded software development process. The speech details will be divided into the following sections
1. What is included in embedded development, including development of driver layer and application layer.
2. What Python can do in embedded development. Python can be used in embedded development of application layer, including background web development, instant communication development, GUI development, etc.
3. Compare the advantages of Python development over traditional embedded development.
4. Briefly explain the basic process of embedded development using Python, and mainly explain the construction process of Python development environment.
5. Finally, showing the specific application intelligent parking system management display and PLC communication read and write operation display.

Working as development manager in iRain Parking, also worked in Hangzhou Hikvision. After working, he has been engaged in embedded software development, mainly involving camera sensor driver development, intelligent access control system development and intelligent parking lot software development.

Building a website is a very fulfilling thing. In the eyes of most people who are interested in driving development, the simple use of Python and its components is complicated by the fact that each module is built into a complete set of solutions. Now that OAuth2.0 and OIDC are standard, is it useful to integrate an identity authentication module into a website? AWS provides some good ideas on development costs, development cycles, and controllability. So, how should we use building blocks to build web applications step by step? Stay tuned.

She has provided consulting and delivery services for microservices and website construction at home and abroad. I am good at testing drive development and software architecture evolution. Passionate about technology, passionate about agile practices, and a practitioner of continuous delivery.

Briefly introduce modules and functions that involve cryptography in Python

Beijing University of Posts and Telecommunications - Network Information Security Winter Camp - Senior Lecturer, Guangzhou University - Network Information Security Summer Camp - Senior Lecturer, Domestic Famous Information Security Research Team - Core Member of Dawning Team, Beijing Mingyuan University - Network Space Advanced Technology College - Senior Lecturer, Freebuf-DigApis Information Security Column Leader

Add extra protection to yourself by means of fixed investment
1.What is the automatic investment plan (AIP)?
2.Advantages of the automatic investment plan (AIP)
3.Shortcomings of the automatic investment plan (AIP)
4.Advantage about programmers do the automatic investment plan (AIP)
5.How to make automatic investment plan (AIP) with Python

He joined IT in 2000 and has been engaged in communication/network security for a long time. He has passed the qualifications of well-known foreign companies, also participated in the entrepreneurial boom. He is working on an information security company currently. In his spare time, he is committed to popularizing Python programming, making Python a tool that everyone can use, organizing multiple Python public interest qq groups, and spreading introductory courses through live broadcast. H e is interested in quantitative investment, and experienced losses in stocks, futures, cryptocurrencies and other varieties, and focused on fund investment, risk reduction, and stable income gradually.

AI is math. Though we do not have to prove something mathematically when creating a deep neuro network, we are using statistical methods to learn the patterns from the data. Unsurprisingly, we've done some unseen assumption behind the model that will never lead to the result that an intelligence could understand and generalize about the world. Today we are going to discuss about the limitations of machine intelligence based on some real world projects, from image recognition to natural language processing, about both data and models. 

To python players, this talk may answer the following questions:
1. How to know whether my problem can be solved by Artificial Intelligence today? 
2. How to design the indictors for a real-life problem? 
3. How to know if a model is 'good enough'? 
4. How much data is enough to train a model? 
5. Tips and tricks when using Python for deep learning. 

Wang Xiaolei, the chief algorithmic scientist of the China Consulting team, has provided data and algorithm related consulting services in many domestic and overseas projects for a lot of companies, including banks and insurance companies, financial services institutions, real estate advertising companies, consumer goods companies, educational service organizations, and communication technology companies. It mainly includes mathematical modeling and optimization, unstructured data processing, natural language processing, image recognition, etc. It has end-to-end experience in human-computer interaction and precision marketing, and truly realizes the value of customized mining data.

1. Introduction to PyTorch
Introducing the characteristics of PyTorch dynamic graphs, and comparing PyTorch with other deep learning frameworks.
2. Introduction to new features of PyTorch (mainly for version 0.4)
Introducing the new features of PyTorch 0.4 (which may include 1.0 if 1.0 is released) to help the programmers migrate the code to the new version.
3. How to build a simple neural network using PyTorch
Introducing the simple use of PyTorch (use of nn.Module)
4. PyTorch's parallel training
Describing how to use the multi-GPU to train the PyTorch model (using the apex library)
5. Deploy PyTorch model with ONNX
Describing how to solidify the model with the ONNX format and deploy the model on Caffe2
6.Tips on PyTorch
Introducing the various pits that you may encounter when writing PyTorch
7. Summary

Now working as a deep learning architecture engineer at NVIDIA, he is responsible for the optimization of GPU-based deep learning models. The main application scenario of the Python at work is to build the deep learning model. The main deep learning frameworks used are PyTorch and TensorFlow. Hope to share the experience on PyTorch's with everyone in PyCon.

The goal of using Python to implement Ethereum from scratch is to disassemble the Ethereum "rocket", to find out, and then reassemble it with Python. Our focus is not on recreating an Ethereum, but on getting an experience in the process. This series can be roughly divided into four chapters, namely P2P network protocol, blockchain structure, consensus calculation and the smart contract. This topic explores the P2P network protocol component, and we will try to connect to the Ethereum network and communicate smoothly in python.

Continuous entrepreneurs, more than 10 years experience in Web application architecture, mainly writing Python/Rust/JavaScript/Haskell language, Ethereum early fans and evangelists, open source software supporters. The current focus is on blockchain payment projects. Yield.IM

Theme introduction:The new asset types (digital currency and vouchers) brought about by the blockchain technology revolution, as well as its decentralized infrastructure, are bringing dramatic changes to the global asset management industry. With its mature large-scale financial application background and active user community, Python language has become the preferred solution in the blockchain asset management business. This presentation aims to share the application of Python in the various ecological aspects of blockchain asset management and to explore future directions.

Founder of Shanghai Wayne Kailuan Software Technology Co., Ltd., ever served as a quantitative trader and fund manager in large-scale quantitative private equity funds, developed the vn.py quantitative trading framework, which is currently one of the world's largest open source quantitative trading projects ( Github Star 6071). The current work focus is on the development of quantitative trading systems in the blockchain and digital currency sectors.

1. Introduction to type system, comparison of static type and dynamic type
2. What is Type Hints
3. Why do I need to use Type Hints?
4.Some ways to use Type Hints
5. Some problems with Type Hints and the improvements in Python 3.7
6. How to start with Type Hints in a real project
7. Continuous integration of mypy for type checking
8. The future of Type HInts, CPython runtime optimization? JIT?

Now working as a backend development engineer (Python) at iQiyi, 4 years of Python coding experience, and mainly use Python for web development at work. In this PyCon, Mr. Zhang will share some experience of using Type HInts in real projects.

Jupyter (IPython Notebook) is cool and popular. Except for All-takenism, it seems difficult to extend Jupyter? This topic is an easy-to-understand example of how to extend Jupyter to link your cloud services, optimize the output of services, and make Jupyter smarter for your projects and customers.

Worked as an architect in Splunk (the first purely data-listed company on NASDAQ in the US) and being in this field for 10+ years, Mr. Ding has a wide range of technical interests. He shared 5 well received topics at the PyCon 2015/2016 conference, covering the Big data analysis and Processing visualization, Python functional programming / design pattern, Python big data scheduling engine, Pythin core injection technology and other topics, etc.

Why is the length of the object being len() instead of obj.length() ? What is the difference between an iterator and an iterable object? Why do we rarely use design patterns when writing Python code? Behind this it is the idea of "protocol" in the Python language. This talk will discuss how to understand the protocol in Python, what is the design idea, and how to write Python code to implement this idea.

- Python multi-version, multi-environment coexistence analysis
- traditional solutions
- Pyenv advantage
- Did you install Pyenv right?
- Best practices of Pyenv

The web name of the speechmaker is ITXiaoPang. He's the founder of the open source sky website. He has been programming since 2003.He is familiar with many languages, but loves Python only. Currently working on Python's operation and maintenance in a well-known game company.

The topic of my speech is about the comparison between a non-Pythonic way and a Pythonic way of approaching business problems. As I switched from a business side to a technical field, I have experienced how two sides differ in solving business problems. I will cover the actual problems needed to be solved on FC Network Analysis using K-Means and how pro Excel users would approach this problem.

Work as a Data Analyst at an SCM Data Science team at an ecommerce company. The team analyze everything related to SCM plus some of FC related topics. Getting to use and learning about Python on a daily basis has been a great experience!

Because of the dynamic type, Python's performance has always been a headache. The release of Python3 did not bring performance improvements. The performance of pypy's interpreter that supports jit is still not good enough. This presentation will use a performance test code to show the optimization step by step and show the tuning process, explaining how to optimize the performance of Python code with numba and cython, as fast as Go.

Responsible for web backend development in the Mozy group, Mr. Jiang has many Experience in the development and optimization of high-traffic Web backends for many years.

江湖人称大妈的大妈 Pythoneer ;-)

Python 中文社区创始人 / 管理员,热心于python社区的公益事业,大家熟知的社区"大妈";OBP及蟒营工程设计者 /主持人;参与并主持各种线上 / 线下活动,主持编撰了 "可爱的Python";坚持用 Pythonic 感化国人进入 FLOSS世界进行学习 / 分享 / 创造!

- Python 在软件测试领域的生态 - HttpRunner 的诞生背景 - HttpRunner 的开发设计思路 - 基于 HttpRunner 的自动化测试方案 - 基于 HttpRunner 的成功实践案例

李隆（debugtalk），DebugTalk 博主，HttpRunner 作者。7年软件测试行业经历，重度 Python 爱好者，始终致力于 Python 在软件测试领域的应用和推广。信奉 Python 之禅，追求代码的简洁和优雅，于人生亦然。当前在大疆创新负责测试开发方面的工作。

近年来，应用形态逐渐朝着 PWA、小程序等范客户端形态发展；云计算则朝着 BaaS + FaaS 的 Serverless 无服务架构大步迈进。本次分享向大家介绍爱范儿团队在用 Python 打造小程序领域的 Serverless 平台知晓云的过程中遇到的挑战和收获，以及对接下来的云计算形态的展望。

爱范儿 CTO、知晓云负责人，大学开始创业，先后参与过多个创新创业项目。资深移动互联网应用专家，擅长架构设计，专注云计算、机器学习及移动开发。

由于种种原因，很多时候说到微服务貌似使用的语言只有 Java，其实微服务的初衷是希望不同的服务能够根选择自己擅长或者喜欢的语言来开发。这个话题就想聊聊用 Python 怎么微服务，主要内容有为什么要使用 Python 来做微服务，使用 Python 如何设计一个微服务架构，有了架构在实现的时候如何编写 Python 微服务，最后简单介绍一下在部署微服务的时候应该注意些什么。希望通过这个话题能够让大家能够建立一个 Python 微服务的体系，在某些场景下能够利用 Python 的优势快速实现基于微服务架构的应用。

十多年IT从业经验，从技术到金融，从金融到风控，从风控到互联网，从互联网到咨询。不断走出舒适区，体验多维的人生。

通过分享在安卓上移植编译 Python 的实践经验与大家一起深入 Python 这个软件工程的结构设计，以及如何阅读 Python 源码，甚至扩展 Python 代码。 涵盖的内容包括 - 在安卓平台运行 Python 的意义 - 安卓平台与标准 Linux 平台的一些共同点和差异 - 编译工具链的方案及一些问题 - Python 代码结构 - 怎么运营一个基于编程工具的社区

优趣天下创始人，全球最流行的安卓 Python 应用 QPython 作者，前Zynga高级开发工程师、前绿盟科技产品经理、前新浪邮件高级开发工程师。

以一个 Pythoneer 的角度见证一个电商5年的沉浮发展，分享在此过程中如何利用各种开源方案解决性能优化，如何在不断变化的业务与不断调整的商业模型中为公司提供高效的工具让其处于有利的竞争地位。探讨80后程序员在不断发展的形势面前如何保持自己的竞争力，如何组建自己的团队，如何健康的生活。

2009年毕业于哈尔滨工程大学硕士学位。曾就职于诺基亚，华为等公司，2013年初加入深圳花果山科技有限公司工作至今。爱好编程，中医，传统武术等。

- 使用 Python 开发并实现Cardano（区块链）项目。 - 对比 Python 和 Haskell 语言在区块链项目方面的特点。 - 介绍 Cardano 项目本身的开发细节。

10多年的老 Pythoneer，从事过 Web 前后端，手游前后端，游戏引擎开发，杂七杂八的编程语言，从码农到 CTO。近年来深入研究区块链相关技术。

深度学习框架下维护聊天助手个性一致的技术框架，很明显还可以衍生出很多种其他方案，但是技术思路应该是类似的，核心思想是把聊天助手的个性信息在Decoder 阶段能够体现出来，以达到维护个性一致的目的。

斯坦福大学计算机科学专业，虚拟化和云计算专家，拥有多项云计算研发专利。曾任职全球数据存储行业龙头企业EMC技术总监等

在异步编程领域 Python 有着十几年的经验和积累, 从 asyncore, futures 到 tasks, coroutines, Python 3.x 的异步能力围绕着 asnycio 在近几年内迅猛提升. 本演讲主题从一个极小的疑问发起, 浅谈协程与事件循环在 Python 中的联系与实践。

Python 爱好者, 关注系统编程领域. 14 年蟒营接触 Python 至今, 本科毕业第一个月的新工程狮一枚, 珠海 GDG 活跃分子.

...

...

...

...

...

1) GraalVM简介及Python在OpenJDK开发中的重要作用 2) 深入分析GraalPython的实现 3) 重新思考和设计Python运行时 4) 在开源ARM平台（如树莓派/96Boards等）上实践GraalPython/GraalVM

先后就职于Motorola, Samsung, Alibaba, EMC, Citrix等IT公司，现为自由职业者。熟悉基于ARM+Linux的嵌入式软件平台架构, 在甲骨文JavaME, 高通BREW, 摩托罗拉P2K，谷歌Android，三星Tizen等移动平台上具备十余年研发经验。 近几年主要从事云计算相关研发工作，在数据中心软硬件架构，软件定义存储，虚拟化，信息安全，高性能网络等技术领域积累了一定的经验。 是《灰帽黑客 第4版：正义黑客的道德规范、渗透测试、攻击方法和漏洞分析技术 》一书中文版的主要译者： http://product.dangdang.com/23927890.html 对技术创新具有浓厚的兴趣和实践能力，热心参与开源社区的各种活动，多次参加各种IT会议并作技术分享

分享应用机器学习对海量已分析的软件运行产生的log进行学习，实现对未分析log进行机器自动分析归类，提高调查问题的效率的解决方案以及技术架构，涉及大数据带来的数据存储，计算机内存，特征提取效率，模型训练效率的挑战。

方案基于诺基亚Cloud，S3以及GPU等资源，应用了spark，tensorflow等framework，以python为主要开发语言，分享使用python在机器学习工程处理大数据的思考与挑战。

主要从事机器学习/深度学习的建模，应用机器学习/深度学习分析公司软件产生的log，提高软件产品开发维护的效率。并任deeplearning.ai课程Sequence Models助教。

基于矩阵的统一计算框架 Mars，目标把 PyData 的生态带到分布式执行环境，目前支持 Numpy 接口，能在分布式环境执行多维数组操作，它支持在 GPU 上运行，且支持二维稀疏矩阵。目前，Mars 兼容 70% 常见 Numpy 接口。本次分享会介绍目前我们的工作，会介绍 Mars 的实现原理，并且对未来的上层建筑包括 DataFrame 和机器学习进行展望。

主要工作是将 Python 生态和大数据环境进行结合。过去工作包含编写了 pyodps DataFrame 框架，这个框架能让用户编写类似 pandas 的代码，但支持在 MaxCompute 大数据平台、pandas 和传统数据库上执行。现在的工作主要是 Mars，这是一个基于矩阵的统一计算框架，目标把 PyData 的生态带到分布式执行环境，目前支持 Numpy 接口，能在分布式环境执行多维数组操作，它支持在 GPU 上运行，且支持二维稀疏矩阵。未来会将 pandas 和 scikit-learn 分布式化。

自然语言处理（Natural Language Processing, NLP）作为人工智能领域的一个重要分支，现在已经成为人工智能研究、创业和落地的热点方向。本次分享将从理论和实践结合的角度，介绍 NLP 的理论、实现和流程，同时将介绍 Python 环境下重要的工业级开源框架：SpaCy 和 RASA NLU，并提供针对中文场景的具体建议。

从事自然语言处理两年多，有丰富的理论和实践经验。目前在吉利集团旗下车联网公司从事车载人机交互系统的研发工作，曾在阿里巴巴集团从事自然语言相关的框架和算法开发工作。

在对外发行 Python 应用时，我们可能出于个人或商业的保密要求，需要加密源码。主流的加密方式有发行 .pyc 文件、混淆源码、使用 Cython 编译等。本次分享将简单介绍现有加密方式的优点与不足，再重点介绍如何基于对称加密和非对称加密算法定制 Python 解释器，从而达到加解密源码的目的。

开源图书《Python 最佳实践指南》中文版译者、开源Python跨平台工具库“pydu”作者，Python热爱者和工作者。从事过采控平台设计、Agent开发、测试工具开发和项目管理等工作。