Interviewer : Ma Wei, SegmentFault Technology Editor

Respondent : Hu Xiaoman, Huawei MindSpore Operations Director, LF AI & DATA Outreach Committee Chair

Hu Xiaoman, Huawei MindSpore operation expert, chairman of the Outreach Committee of the LF AI & DATA Foundation, and member of the expert database of the Chinese Institute of Electronics. The technical leader of the TinyMS project and the founder of MSG·Women In Tech. He used to be the head of the algorithm and operation team of Autohome, Baidu and other companies, responsible for multiple deep learning implementation projects and open source project community operations, and was selected as the "2021 China Open Source Pioneer 33" annual list.


https://www.bilibili.com/video/BV15g411T72Q/?aid=506673615&cid=442643026&page=1

A product manager who doesn't understand technology is not a good open source operation

In 2015, Hu Xiaoman, who had just graduated from the Department of Mathematics, became an algorithm engineer. While writing code, she wrote the principles of machine learning algorithms on a blog website. The easy-to-understand method quickly made her well-known in the technical circle.

Since 2017, Hu Xiaoman has written a technical tutorial of "Introduction to Deep Learning from Zero in Three Months", and quickly became a well-known technology blogger in the blog garden, with over 5 million hits. At the same time, he was also invited to start various online blogs from time to time. Share below.

In 2019, Hu Xiaoman went a step further on the road of technology sharing and became a "deep learning evangelist", helping more developers quickly enter the door of the deep learning world. During this process, Hu Xiaoman became attached to Huawei and joined Huawei in 2020 to be responsible for the operation and management of the open source community of MindSpore, Huawei's all-scenario AI framework.

"At Huawei, I can think about how an open source community grows from 0 to 1 from a holistic perspective."

Different from the role of technical evangelist, the most attractive thing about Hu Xiaoman's career at Huawei is her ability to participate in the growth of an emerging open source community from the very beginning. For Hu Xiaoman himself, it is also a new breakthrough from the position of programmer to community operation. It is indeed a rare and precious experience for a cross-border person to fade away from youth and mature with a young community.

understand technology, but also understand users

Today, the MindSpore community has more than 800,000 users, 4,000 contributors, and more than 1 million downloads. Leading the open source community to such a prosperous scene from 0 to 1, Hu Xiaoman summed up the following experiences:

1. Solid technical foundation

Hu Xiaoman, an algorithm expert, has a solid technical foundation. It is quite easy for Hu Xiaoman to get started and use the MindSpore framework, and she can also deeply participate in and understand various technical topics discussed by developers. In her opinion, the operation of the open source technology community should at least be able to independently use the operating products, and it is best to have a deep understanding of the content related to the code and principles. With such a technical background, operations can better integrate with community users, experience the community atmosphere, and understand the needs and pain points of users.

2. Diversified operation methods

Hu Xiaoman believes that the operation of the open source community needs to be combined with the development cycle of the project. In the early stage of open source, she set the goal to "rapidly increase the technical influence". MindSpore is an AI framework with a very fast iteration speed. In the early stage of open source, a new version will be released at the end of each month to allow developers to experience the new version features. How to let developers understand the advantages of these new features faster is her most important One of the first problems to solve.

In the process of continuous attempts, Hu Xiaoman integrated the successful experience of various Internet communities, so he began to use videos, interesting courses and other methods to attract users, and achieved quite good results. Technical videos are often boring. Hu Xiaoman combined the pain points of developers with the features of the new version, shot a 1-minute short video of the features of the new version, and posted them on Douyin, Station B and other websites. In the absence of any channel promotion, the number of broadcasts on the entire network exceeded one million, and the technology brand of MindSpore was quickly launched. She came from an algorithm expert background. She applied her previous model training experience to open source operations, and for the first time proposed the methodology of "operating with model thinking", taking each operation method as a small model and following the entire "model" Iterative optimization is the same as the “development” process. After the launch, the evaluation of the effect, the collection of community feedback, and the next improvement are eliminated. This evolutionary method has become one of the main driving forces for the rapid growth of the MindSpore community.

3. Equal and friendly community atmosphere

In addition to technical background and diversified means, Hu Xiaoman mentioned the most critical value points in the construction of open source communities.

"In a real open source community, operators and developers should be on an equal footing."

Many community managers of open source technologies have a condescending psychological cognition, believing that developers are only "contributors", placing themselves in the dominant position of cooperation, and even having a parental mentality. However, Hu Xiaoman believes that the status of community operations and users must be equal and open, so that users can feel the value of their participation and contribution.

This concept of equality and openness runs through the development process of the MindSpore community, and has been recognized and supported by the majority of community members. Many developers will continue to put forward their own suggestions, and operators will give corresponding feedback soon. Such exchanges promote the growth of the entire community. Such a community atmosphere also gave Hu Xiaoman a greater sense of achievement, allowing her to realize that her work has truly brought benefits to thousands of developers and community members.

It is difficult to quantify the success of an open source community. The open source community health index of upstream communities does not have a standard evaluation system, and even if there is, it is difficult to apply to all open source communities. Other platform-specific indicators, such as GitHub's star, fork, watch, etc., are not the only criteria for measuring the success of an open source community, so what kind of target is selected as a consideration indicator determines the future direction and value positioning of this open source community. In this process, it is necessary to repeatedly verify, run-in, and prove to ensure that the positioning of the MindSpore open source community is a community that is completely developer-oriented, not a community that is purely theoretical. Metrics are there to measure one's own work, not the sole purpose of the developer community.

Product thinking is a community, success!

1. ToB New Paradigm: What is MSG·Corporate Behavior that has a higher conversion rate than the traditional model

Whether open source products can bring commercial value to enterprises has always been a question in the industry. The traditional ToB model requires a lot of manpower and time, and the cycle is long and the results are slow. How to understand the pain points and demands of the enterprise more effectively and quickly is the core point that Hu Xiaoman wants to break through. In the second half of 2021, she designed a complete set of MindSpore enterprise expansion plans, which have targeted plans for enterprises of different types, industries and stages. In July of this year, the first MSG Enterprise Tour, Nanjing Field, was launched. It joined with Jiangsu Shengteng Innovation Center, covering 96% of Nanjing’s local AI technology companies, empowering enterprises in advantageous industries, and letting them understand MindSpore’s technological advantages, The support from the computing center and innovation center has built a bridge for cooperation between local small and medium-sized AI enterprises. In the follow-up, the successful model of the enterprise line has been brought to Wuhan, Xi'an and Chengdu. It has established cooperation channels for more than 500 high-intent enterprises, which greatly saves the funds that the enterprises themselves need to spend on AI talents and computing power, which has significant commercial value.

2. Project delivery system: How does TinyMS help the team win the gold medal in the competition within half a year of open source?

At the end of March last year, Hu Xiaoman led the team to develop a new AI tool, TinyMS, a high-level API tool based on MindSpore. At the beginning of the project, Hu Xiaoman was determined to design a high-level AI framework tool for AI beginners that can be used with zero foundation, and equipped with nanny-level tutorials to help Xiaobai from the most basic shell scripts, Start learning python, necessary mathematical knowledge, etc., until you can use the handwritten model yourself. In October, the finals of the 7th China International "Internet +" College Students Innovation and Entrepreneurship Competition opened in Nanchang. The team from South China University of Technology chose the TinyMS competition question - "Using the MindSpore Advanced Tool TinyMS Supported Network CycleGAN Training Image Style Transfer Model", won the only gold award in the industry track in the fierce competition of more than 12,000 students in 6,100 projects. After interviewing the representatives of the two award-winning students, they both believed that it was precisely because of the extreme ease of use of TinyMS that they could complete the development and debugging in such a short time, and finally won the championship. Hu Xiaoman always believes that in the operation and management of the open source community, for any project, there must be a clear and clear positioning and planning, and a project delivery system should be adopted to meet the principle of minimum MVP first, and then continue to optimize and iterate. To make the product more perfect to meet the needs of users.

3. Ecological Inheritance: The Story Behind the Landing of Two Ecological Cases

In April 2021, in one afternoon, Shanshui Nature Conservation Center and the MindSpore community hit it off and decided to do a big thing for the state to protect animals and cooperate in the Sanjiangyuan field infrared camera identification project. After half a year of repeated tuning and field experiments, it will finally be launched in the Shanshui Nature Conservation Ecosystem by the end of 2021. This is the first pre-training model based on the domestic framework of Sanjiangyuan infrared image nature conservation, and all related tools are open sourced.

In addition to ecological protection, cultural inheritance is also indispensable to the help of MindSpore. In the MSG·Women In Tech event organized by Hu Xiaoman, Ms. Song Ting, an AI and encryption artist, wanted to use AI to contribute to cultural heritage. Based on MindSpore's GAN network, Song Ting's team made a tie-dye tie-dye for the world's protection of intangible cultural heritage. The pattern is trained to generate new tie-dye patterns to protect non-inherited craftsmanship, and finally broadcast on CCTV as an open source cultural feature case.

Many people in non-AI fields often don’t know what changes AI can bring to people’s lives and what contributions it can make to nature. These two cases vividly demonstrate the impact of the MindSpore AI framework on the natural ecology and cultural fields. This is also Hu Xiaoman's view that investing in open source is the greatest value.

Want to get in? do this

How to build yourself into an excellent community operation expert? Hu Xiaoman reviewed his own experience and put forward three points for your reference.

1. Keep the original intention and empathize with the developers

When you enter a new field, you will definitely feel at a loss, but after getting familiar with it, it is difficult to understand the state of mind as a Xiaobai. So keep yourself fresh at all times, empathize with the developers, and review the mood at that time. In addition to her daily work, Hu Xiaoman will often learn different technical frameworks, such as Julia, etc., to understand the development history of other different types of technical communities, the update of technical features, and the construction of community atmosphere, which can help her jump out of the original AI. In the field, look at the original work from another perspective, whether there are points worth learning. In addition, in order to maintain the same frequency as the developers, Hu Xiaoman will regularly chat with the developers to understand their experience when using the product, and truly face the developers at zero distance.

2. Broadly dabble, quickly transform inspiration and execute

Hu Xiaoman has a very wide range of reading. In addition to the reference books he used to read when he was a programmer in the early days, he also covered various types such as philosophy, natural science, and social psychology, and recorded points worth thinking about in his reading notes. . When reading books in different industries and fields, she often combines it with the application scenarios in her work, leads to some new thinking, then converts the thinking into an actionable execution plan, then quickly iterates, and finally abstracts it into a feasible solution. The methodology will be reused in similar scenarios in the future. In addition, she will also experience various Internet communities, including Xiaohongshu, Douyin and other content platforms, to understand how each platform attracts users to spontaneously generate content, distribute traffic, set up incentive mechanisms, etc. Although they are different products, successful products must have commonalities and can summarize some common experiences, especially in terms of grasping human nature. These full-category content platforms use this to the extreme. For open source communities, in addition to the strength of the technical products themselves, a good content attraction mechanism is also one of the essential factors for community operations.

3. Quantitative indicators to continuously improve work effects

If you try a new thing, without any goals and deadlines, it's easy to procrastinate over time, or give up because you don't see progress for a long time. Therefore, Hu Xiaoman proposed that when dealing with work or new things, we must learn to quantify learning, concretize indicators, complete quantitative completion, and update stages. Why come to this conclusion? In fact, in daily life, Hu Xiaoman loves fitness and actively tries various sports, including badminton, Pilates, skiing, etc. After studying a variety of sports, she found that fitness, or different sports, can be quantified. of. For example, in fitness, many people will be easily injured if they directly lift heavy weights without systematic learning and scientific methods, and they have no fixed goals. Every time they study in general, they will feel frustrated in the long run. In the same way, when developers come into contact with a new field and new framework, if there is no periodic feedback, it is difficult to find their own problems, and people are often lazy, so Hu Xiaoman designed a development in the community. The developer advanced system allows all developers who enter the MindSpore open source community to clearly find their own position, guide and motivate developers to continue to move forward, so that every developer can continue to grow in the community. In his personal work, Hu Xiaoman often quantifies his work indicators, and uses data to quantitatively feedback the results, instead of using experience to describe the quality of the results.

Beyond the mainstream, more attention should be paid to

1. Focus on female developers: perspectives and insights from women

Planning an offline salon event on the theme of female developers is a unique achievement of Hu Xiaoman's operation work. She is the first person in China to pay special attention to female developers in the open source community. Hu Xiaoman, who is a female programmer, is more likely to understand the psychological cognition and worries that women are a minority in the IT industry. For example, many women will have various doubts about their career prospects and career development, so Hu Xiaoman specially planned MSG·Women In Tech's women's special event, aiming at female practitioners in the field of technology, and forming a unique space for them, Let the female seniors in the industry share their own cases and solve the confusion in the development of the workplace.

Common technology salons are often sermon-style, with several experts giving speeches and attendees listening carefully, with only a small percentage of interactive parts. But MSG·Women in Tech is no longer just listening, but fully interacting with the guests and discussing career issues with the participants. So far, a total of 4 events have been held, and more than 20 outstanding female industry seniors have been invited to share their experiences in Beijing, Shanghai and Shenzhen, attracting 300+ female technology practitioners to sign up, and 100+ women came to the scene to communicate with each other . Share a series of common problems in the workplace, such as how to become a qualified technical leader, how to manage upward, how to strive for the projects you want to do, etc. The participants and guests sat in a circle around the table, and everyone went to explore solutions and share Experience, in September last year, as the official representative of Huawei Women In Tech, the shooting video was released on Huawei's official account at home and abroad, and the spread exceeded 10 million, which was widely recognized.

2. Open source operation is also open source: it is not just code that can be open sourced

In the open source community, common open source projects are usually code-based, but Hu Xiaoman thought, why can't we open source operations as well? Not only code experience can be reused, but the open source community operation itself can also be summarized into experience to form a methodology and then open sourced to benefit new students. At the end of March last year, the MindSpore open source community and the Open Atomic Foundation jointly launched an open source community 0xCommops that "opens up community operations as well". Tool Library.

some small suggestions

Many young technical practitioners and even young people with non-technical backgrounds are interested in positions such as open source community operation, and Hu Xiaoman also made valuable suggestions for these small partners.

  • Learn about the industry . Hu Xiaoman suggested that in order to make achievements in the post of open source community operation, the operator must first have a very solid technical background and enough understanding of the products they want to operate. A qualified open source community operator must be a senior developer with product thinking.
  • quick trial and error . Operation work does not have to be done by entering a company. If you are a programmer, you can start by operating an open source project of your own, try to be an active community from 0 to 1, summarize effective experience, and take small steps. Going fast, fast trial and error, is very helpful for officially entering the work of open source operation.
  • maintain great enthusiasm and abundant energy . It is a very rare thing to engage in a career you love. If it happens to be, you are lucky, keep your original intention and enthusiasm, if not, turn it into a job you love, find the excitement in your work, and so on to go further. There are many possibilities for the career development of operations, and it is necessary to constantly think and explore in the work, and modify the career picture according to oneself. With the continuous expansion of their own horizons, more achievements, and continuous accumulation of knowledge and experience, operators will also have a clearer understanding of their own strengths and weaknesses. On this basis, small partners can choose the most suitable development direction based on their own interests.

All in all, love your work, treat it as a career rather than a career, and go on firmly will make you go further and bring unexpected value!


About Huawei Open Source

As a trusted open source citizen, Huawei works with partners to promote inclusiveness, fairness, openness, and more solidarity through continuous contributions, build a world-class basic software open source community, and accelerate the digitalization of the industry.

· Active participants and supporters of mainstream open source organizations. At present, Huawei is a top/start-up member of dozens of international open source foundations.

· Contribute to open source basic software on a large scale, consolidate the ecological foundation of digital infrastructure, and work with partners and developers to build an open source ecosystem. In the past two years, Huawei has open sourced multiple platform-level basic software open source projects such as KubeEdge, MindSpore, openEuler, openGauss, and OpenHarmony for cloud native, automation, and intelligence, and has become an open source community accepted by developers around the world. Industry commercial landing.

· Actively build a sustainable, viable and credible open source community. Huawei is committed to improving the community ecological governance structure to ensure the continuous evolution of the community.

Pay attention to Huawei's open source public account to learn more!

Click here to enter Huawei Open Source official website to learn more


思否编辑部
4.4k 声望117k 粉丝

思否编辑部官方账号,欢迎私信投稿、提供线索、沟通反馈。