2010: The Tech Groups Inside Baidu Hi
· 25 min read · Views --

2010: The Tech Groups Inside Baidu Hi

Author: Alex Xiang

Essays

In 2010, I joined Baidu.

That sentence looks calm now, but for me at the time it was a major turn. For a long time before that, I had been working around Linux, system software, file systems, and cluster management. Those jobs were close to machines and far from ordinary users.

Baidu was different.

Search is an extremely internet-native system: many users, many requests, massive data, fast change, and every technical decision forced back into the context of scale. In the past I cared about one machine, one file system, or one cluster. After joining Baidu, I had to care about traffic, latency, recall, ranking, logs, release, monitoring, and the behavior produced by a large crowd of users every second.

The Person Who Opened The Door

I have to write this part down.

I was able to enter Baidu because of the leader who hired me at the time, Liancheng. He was a very good person, and he gave me the opportunity to move from the route of foreign companies and system software into the internet industry.

Many career turns look, in hindsight, like choices we made ourselves. But whether we could pass through that door often depended on whether someone was willing to believe we could adapt to a new scenario.

Moving from system software to the internet did not mean all skills became irrelevant. C/C++, Linux, performance work, debugging, scripting, and service stability all transferred. But internet companies had their own rhythm: product changes were faster, user feedback was more direct, data-driven decisions were stronger, and engineering boundaries were more complicated.

Without that opportunity, I might have continued along the system-software, storage, virtualization, and foreign-company path. That might not have been bad, but the next dozen years would have been a completely different story.

The Tech Groups Inside Baidu Hi

Baidu still had a strong engineering atmosphere then.

The internal IM tool was Baidu Hi. Hi had many technical groups: C++, Python, PHP, Linux, system problems, online incidents, performance optimization, utility scripts. The discussions were lively.

I have always felt that a company’s engineering culture cannot be judged only by slogans or process documents. The more real things hide in daily communication: whether engineers are willing to discuss technical problems, whether senior people are willing to answer, whether disagreement is taken seriously, and whether incident reviews look for causes instead of scapegoats.

The Baidu Hi technical groups at that time felt alive. You could see many people genuinely solving problems, and you could see many people still interested in technology itself.

Search engineering transition

During those years, Baidu was still a place many engineers wanted to join. It had enough traffic and enough system complexity. Search, advertising, crawling, anti-spam, natural-language processing, distributed systems, data processing, and online services could each go very deep.

From Machine View To User View

After joining Baidu, I mainly used C/C++, Python, and PHP.

That language mix represented the internet engineering scene of the time quite well. C/C++ handled performance-sensitive and lower-level services. Python handled scripts, tools, data, and automation. PHP handled many web and business interfaces. It may look mixed today, but that was the real state of the field.

The biggest change for me was the perspective from which I looked at systems.

When I worked on file systems and cluster management, I mostly started from machines, processes, files, networks, and configuration. In search and internet products, machines still mattered, but everything eventually had to be mapped back to user behavior: what users searched, what they clicked, how long they waited, whether they retried after failure, and how many people were affected by a system wobble.

Technical metrics were no longer only engineers’ own metrics. They mapped directly to business and user experience.

This later made me understand the importance of data more deeply. Without data, many arguments are only feelings. With data, problems remain complicated, but at least the discussion can start from facts.

Entering The Golden Age Of The Chinese Internet

Baidu in 2010 was still at a very high point.

In public financial reports, Baidu’s revenue in the first quarter of 2010 grew by nearly 60% year over year, and the second quarter grew by more than 70%. Behind those numbers were search advertising, the growth of Chinese-language internet content, the expansion of internet users, and businesses moving marketing online.

Also in 2010, companies and products that later became very important began to appear, including Meituan, Xiaomi, and Zhihu. Meituan started from group buying and moved into local services. Xiaomi combined phones, MIUI, hardware, and internet services. Zhihu entered through a question-and-answer community and knowledge content. When I had just entered Baidu and stood inside a mature search giant, it was hard to immediately realize that another group of companies outside was forming new entry points.

Inside the company, engineers did not read financial reports every day, but growth could be felt. Growth brought resources, and it brought opportunities. If a system could not hold up, it had to be expanded. If a tool was hard to use, someone built a better one. If a process was slow, someone changed it. If a direction had value, a team could form around it.

That was the most attractive part of the internet’s golden age: the problems were large, the systems were complex, and companies were still willing to invest.

Of course, a golden age does not stay golden forever. Organizations grow larger, processes become heavier, businesses hit ceilings, and cultures change. But when I entered Baidu in 2010, what I saw was still a large company with a strong technical atmosphere, fast business growth, and space for engineers to make an impact.

The next year, that feeling would become even stronger. Baidu felt like the “Whampoa Military Academy” of that era’s Chinese internet. Many people started from there and later spread across different corners of the industry.

IT Events Of 2010

  • Mobile internet kept heating up. The competition between iPhone and Android entered a faster rhythm. Smartphones moved from high-end novelty toward the center of the next platform war. Applications, servers, and data systems began reorganizing around the mobile entry point.
  • Baidu kept growing quickly. Baidu saw rapid revenue growth across several quarters in 2010, and search advertising remained one of the strongest business models on the Chinese internet. Search, advertising, crawling, anti-spam, and distributed infrastructure trained many engineers.
  • Meituan was founded. Meituan began with group buying and later moved into food delivery, in-store travel services, instant retail, and local-service infrastructure. It represented one important path in which internet companies moved from information services into offline transactions.
  • Xiaomi was founded. Xiaomi combined phones, MIUI, IoT, and internet services into a new kind of consumer-electronics company. Hardware was no longer only hardware; it became an entry point for software, services, and communities.
  • Zhihu began to incubate. Zhihu entered through Q&A and knowledge content, later becoming an important Chinese-language discussion and knowledge community. It showed that beyond social feeds, professional content and long-form answers could also form a stable community.
  • Large internet companies trained systems engineers. Chinese internet giants began absorbing large numbers of system-software, backend, data, and algorithm engineers. Search, advertising, and infrastructure became training grounds for large-scale internet systems.

References

More in this column

View all posts in "Thirty Years in IT and Me"