Exclusive 丨High Robin Li and Zhang Yiming's "technical ceiling": Liao Ruoxue is involved in the tide of autonomous driving entrepreneurship

To be a champion team, every position needs to be the best player possible.

”

Author | Lu Jieping

Edit | Wu Wenliang

Another Internet technology "god" has devoted himself to the autonomous driving industry.

Leifeng network exclusive news, the former Baidu company technical committee chairman, vice president of AutoNavi, Meicai CTO Liao Ruoxue, joined the autonomous driving company Qianhang Technology, became one of the co-founders.

The name of Liao Ruoxue is not unfamiliar to the technical circle, and Tao Ji, CEO of Qianhang and former general manager of Baidu's autonomous driving division, has described him as a "ceiling" level technology god.

1

"Ceiling-level" tech gods

In 2005, Baidu Search's technical team was less than 20 people.

Liao Ruoxue, who graduated from Peking University and worked in game development for 4 years, after the interview and assessment of Chen Zhenyu, Liu Jianguo and Cui Shanshan, liao Ruoxue successfully joined Baidu in April of Yangliu Ruoxue in Beijing and reported to Chen Zhenyu.

It was an era of heroes, and the young examiners sitting on the interview table at that time had a thick stroke in the history of the Internet: Chen Zhenyu served as the CEO of Qunar, Liu Jianguo once became the first CTO of Baidu, and Cui Shanshan, as the only woman among baidu's seven musketeers, served as a senior vice president at Baidu...

After joining Baidu, Liao Ruoxue, Yang Zhenyuan, Tang Huijun and others often cooperated late at night in a small building near Peking University, creating the "earliest" in the Internet industry: such as the earliest to move the SSD flash memory card to the server, the earliest to land the intelligent optical network, etc.

During his time at Baidu, Liao Ruoxue was first responsible for web search, from algorithm strategy to engineering implementation, and created the Baidu Search Technical Committee, which later grew into the Baidu Technical Committee.

Liao Ruoxue told Leifeng Network that the original intention of the establishment of Baidu's technical committee was mainly to solve human problems, one is to help engineers determine the direction of career development, and the other is to set technical grades for engineers and provide promotion paths, which then became the "hard currency" of the Internet industry for a long time.

Another very important work of the technical committee is to initiate discussions on cutting-edge technologies and products, such as the launch of Baidu's new generation of search engine "Aladdin" platform in 2009, when the participants in the discussion were Xu Dongliang, Hu Song, Chen Jingkai, Li Yanhong and Li Yinan also stood for this many times.

Liao Ruoxue introduced that the initial technical logic of the "Aladdin" project is to use the search engine to directly answer the user's questions, for example, when the user searches for how high Mount Everest is, it directly gives the number, this function uses the most advanced knowledge graph and natural language processing technology at that time, although there are still some long-tail problems at that time limited by technology, but this innovative product performance is still very good, greatly increasing the readability of useful information for users.

Aladdin's layout can more or less see Baidu's foresight in cutting-edge engineering technology. With this foresight, in 2011, Liao Ruoxue set up a Baidu recommendation and personalization department to promote the landing of Baidu's internal recommendation and personalized search, and ByteDance, which relied on personalized recommendation to kill the four parties, was not yet born. Even the information flow of one of Baidu's most successful products in the past decade has seen the shadow of what Liao Ruoxue's team planned and envisioned at that time.

After working in Baidu for 9 years, the tide of mobile Internet hit, so that Liao Ruoxue, who was sincere about achieving from 0 to 1 in technology, saw a new direction, and after being introduced by former colleagues, after chatting with AutoNavi CEO Cheng Congwu about the future direction of mobile Internet, he decided to join AutoNavi as vice president of technology and report to Cheng Congwu.

In the view of Cheng and Liao, the map is a natural gene that has become a leading product of local life, in layman's terms, it is to recreate a METUan with a natural traffic entrance. This idea, to this day, nearly 10 years later, still coincides with Yu Yongfu's vision of AutoNavi after taking over Ali's local life last year.

At that time, the Internet industry as a whole was actively engaged in the mobile terminal, in the direction of business transformation of AutoNavi, into the idea of congwu, Liao Ruoxue and other management teams, one is to plan based on maps, do infrastructure APP, somewhat similar to today's local life METUAN APP concept, the second is to make the map production mode from the regular production mode that is updated every few months to a real-time rapid production mode, in a more automated way to process map-related data.

"Then there are things to do pattern recognition, such as directly identifying the speed limit in the road sign line, the left and right turn prompts, and so on." As the technical leader of AutoNavi at that time, Liao Ruoxue joined, making the product of AutoNavi map quickly gain a qualitative leap in technology.

After acquiring AutoNavi in Alibaba in 2014 and completing his phased technical mission, Liao Ruoxue submitted a resignation to Yu Yongfu and Daniel Zhang, and under the introduction of friends, met Zhang Yiming at a café near Soho in Wangjing, Beijing.

The two were very happy to see each other, and their views on personalized recommendations resonated quite well, and even after many years, Liao Ruoxue still remembered Zhang Yiming's rather rational sentence at that time:

"If the problem is solvable, don't let emotions and surroundings sway your rationality, just empty everything to find the best solution." 」

This top technical expert, who still has to write two or three hours of code every day even in a management position, was bound by various personnel problems before, and after listening to the extremely rational Zhang Yiming's suggestion, Liao Ruoxue suddenly became enlightened, and this sentence also became a decisive factor for Ruoxue to join byte.

During Byte's brief tenure, Liao Ruoxue was responsible for Byte's entire technical system. After leaving his job for various external reasons, Liao Ruoxue introduced Yang Zhenyuan, who had fought side by side with himself for many years, liked to think deeply as Zhang Yiming, and never forgot to write code on the front line.

According to Liao Ruoxue's recollection, he was traveling in Thailand at the time, so he and Yang Zhenyuan had many transoceanic phone calls and spent a lot of telephone bills. Under the mediation of Liao Ruoxue, Yang Zhenyuan resolutely decided to join Byte.

The three of them have also become a widely circulated story in the recommended engine circle.

In 2018, Liao Ruoxue joined Meicai as CTO, cutting into the fresh supply chain track from many aspects such as business, product, operation and technology, leading the production and research team and front-line business operations, and optimizing the efficiency of the fresh supply chain through products and technologies.

2

"Early interest in autonomous driving"

The reason that attracted Liao Ruoxue to the field of autonomous driving was that he was interested in autonomous driving technology, and the other was that he saw the commercial value in it.

As a person who has always been interested in cutting-edge technology and has to write code for several hours every day, Liao Ruoxue has long been concerned about automatic driving technology, believing that automatic driving is the next technical mother ecology after search, which can drive a lot of specific technological progress, which is a very interesting area.

However, at that time, the overall self-driving technology was still in the exploratory stage and did not reflect commercial value.

"Recently, we can see more and more autonomous driving technology commercialization, it has changed from the exploration period to the growth period, and the combination of technology and commerce has become more interesting."

After joining Qianhang Technology this time, Liao Ruoxue will be mainly responsible for the Infra part, that is, to build efficient and reliable infrastructure for the automatic driving system to improve R&D efficiency, including the optimization of the underlying operating system and software and hardware integration of the car terminal, the very large-scale distributed data storage, the very large-scale model training system based on hybrid cloud, and the interactive simulation platform.

In the commercial operation stage, the infrastructure will also include large-scale vehicle operation platforms and software OTA systems.

Qianhang Technology is an L4 level autonomous driving start-up established in July 2021 for the trunk logistics scenario.

Its first appearance was in October 2021, the joint investment news of the head investment institution (IDG Capital), logistics platform (SF Holdings), the new power car company (Xiaopeng Automobile), and the technology Internet company (Baidu Venture).

At the end of March this year, its external officials announced the former core figure of Baidu autonomous driving, the former general manager of Baidu's autonomous driving business department, the former general manager of Baidu's intelligent transportation products and research and development, Tao Ji, joining as the co-founder and CEO of Qiangu Technology.

Tao Ji said: "The company is a new company, but the team is a veteran."

Its other two co-founders are Sun Haowen, former technical director of Xiaoma Zhixing Truck BU China, and Ding Fei, an investor in the former IDG Capital Technology Travel Direction.

Born in Baidu, Sun Haowen was already a T7-level senior engineer at Baidu, and worked side by side with Tao Ji to participate in the first trial operation of Baidu's autonomous driving fleet in Wuzhen, and later joined Xiaoma Zhixing, becoming the first engineer of China R&D Center (Guangzhou), the core member of the founding team of Xiaoma Zhixing Truck and the technical leader in China.

Ding Fei worked for IDG Capital for 7 years as an executive director, invested in star projects such as Weilai, Xiaopeng, Xiaoma Zhixing, Luminar Tech, ZOOX, etc., and also performed as a BD VP and coO of the truck team, participating in the commercial exploration of autonomous driving.

The three veterans, coupled with Liao Ruoxue who joined this time, from technology to commercialization to team management, the founding team role of Qianwu Technology has been initially established.

3

"It's not a cross-line"

Entering the autonomous driving circle from the Internet industry, Liao Ruoxue does not think that this is a cross-line, and said that the two have a lot of technical similarities.

This is evident from the professional experiences of many of today's self-driving company executives.

Peng Jun, founder of Xiaoma Zhixing, Wang Jing'ao, CTO of Yunji Zhixing, Zhang Bo, CEO of Didi Autonomous Driving, Xia Huaxia, general manager of Meituan's automatic car distribution department, Gu Weihao, CEO of Ma mo Zhixing, etc., from Microsoft or Google or Baidu or Yahoo, they have all experienced the process of shifting from the Internet field to the field of automatic driving.

The reason is that the field of automatic driving is actually one of the specific application scenarios of artificial intelligence technology.

For example, in terms of perception and prediction, in order to obtain the ability to surpass human drivers, deep convolutional neural networks must be used; in terms of decision planning, imitation learning and optimization algorithms will be used to achieve decision-making on complex road conditions; and the use of neural networks and deep reinforcement learning methods is indispensable in control execution.

And these talents are now mostly distributed in artificial intelligence and Internet companies.

Like product personalized recommendation, search engine optimization, accurate advertising delivery, face recognition, image search, speech recognition, automatic photo beautification, machine translation, etc. The standard functions that are now commonplace on various platforms are inseparable from artificial intelligence technology.

Therefore, people who work in "data mining", "algorithm research and development", "engineering development", "basic platform construction" and "computer vision development" in Internet companies are hot in the field of automatic driving.

It is also these talents that continue to pour in from the Internet field that form a continuous beating heart, providing the vitality and vitality of the autonomous driving field forward.

In addition to artificial intelligence technology, Liao Ruoxue said that infrastructure work such as data center and operating system construction in the field of automatic driving is also closely related to the Internet industry, "Their underlying foundation is similar, related to computer science, mathematics, and physics, it is easier to learn from each other." ”

People who can do well in different fields need strong "technical understanding" ability, that is, to be able to grasp the main points and keys of core technologies, and have enough vision to understand and solve problems.

As a start-up company, Liao Ruoxue said that the primary goal of Qianhang Technology is to rapidly iterate through a large number of scenarios, mass production vehicles, and data closed-loop drive autonomous driving technology, and now one is to solve the problem of how to collect effective data, and the other is to improve data mining capabilities, that is, to turn a large amount of effective data into valuable cases or issues.

This is very similar to the idea of rapid iteration of Internet products, but after chatting with many autonomous driving companies, he found that some problems that have been solved in large-scale Internet data platforms are not well solved in the data closed loop of autonomous driving companies today.

In his view, if the vehicle only runs back and forth on the same highway, even if a large amount of data is recovered, because these data are mostly duplicated, it cannot bring greater value to the case library, and if the hardware is updated and changed, how to make the data always meaningful is a problem that needs to be solved.

"Our more important goal now is to improve our real data utilization capabilities, hoping to run 1 million cases or issues every day."

In terms of simulation testing, Qianhang Technology directly benchmarks waymo, the industry leader, with the goal of running 1 million kilometers of simulation tests every day, and plans to achieve it within one or two years, "the sooner the better".

Of course, in addition to artificial intelligence and infrastructure technology, the field of autonomous driving is also inseparable from the control discipline, vehicle engineering and other majors.

Since autonomous driving involves car control and is highly related to safety, this allows Liao Ruoxue to experience a very different way of doing things than the Internet industry.

Internet as a virtual world, can allow developers through rapid iteration, low-cost way to approach the optimal solution, the Internet field has AB test, gray test, often said "no, you try", which provides Internet companies with the soil for high-speed iterative development.

However, due to the great importance attached to safety issues in the field of automatic driving, coupled with the fact that the construction of software infrastructure as a whole is not mature and the hardware iteration cycle is long, "what we often consider is how to approach the optimal solution as soon as possible under the constraints of ensuring safety and hardware conditions."

For example, he said, doing search and doing autonomous driving have to solve a lot of kernel case problems.

In terms of search engines, in the early days, different problems were solved through various rule-based, and automatic driving was the same, "but in the search field, if there is a bug, such as users searching for things that can't be found, they are angrily scolded twice at most, but if the automatic driving field does not really solve the problem, there will be security problems, so we must make full use of the ability of the Internet to quickly iterate, but also fully respect the laws of doing things and developing in the automotive industry."

4

"Forming an All-Star Team"

In fact, Liao Ruoxue has joined Qianhang Technology since the early days of its establishment, when the office of Qianhang Technology has not yet been fully built.

The commercial value of interest and technology are the two major reasons that attract Liao Ruoxue into the field of automatic driving, and the reason why he joined Qianhang Technology lies in the consensus among the founding team members on the landing path of automatic driving technology and their recognition of each other.

"At that time, my friend took me to talk with various related self-driving companies, and I met several other founders, and I found that the goals were quite the same, and I thought it was very good." Liao Ruoxue revealed that after everyone recognized it, they began to work together.

In Liao Ruoxue's view, the most obvious advantage of the team of Qianhang Technology is that it has a very diverse source of talents.

As the autonomous driving technology crosses the demo technology verification period and comes to the commercial landing exploration stage, it will bear more commercial expectations and mix into a more complex technical system, and the industry is bound to conduct a comprehensive assessment of the company and the team from the dimensions of technical maturity, product definition and iteration ability, mass production landing speed, capital support ability, team management ability and other dimensions.

"The mobile Internet industry has been developing for many years, and many problems have actually been deeply understood, so when making technical decisions, planning or architecture construction, we can design it through a top-down approach."

However, in the field of automatic driving, whether it is the technical path or commercial landing and the competitive landscape, it is still constantly updating and changing, far from being fixed, and the fog is full of fog.

Liao Ruoxue described that if you work in an Internet company, you choose the existing wheels, and when you work in an autonomous driving company, you have to build your own wheels, and even if the wheels are round and square, you need several wheels, you have to discuss and argue.

From how many to install on the car, where to install, what kind of sensor layout to install, and the details to whether to adapt middleware for different hardware, these are still no unique, optimal solutions.

This requires a leadership team with sufficient depth and breadth of experience coverage and a high degree of alignment of values and goals to lead everyone.

Liao Ruoxue believes that the members of Qianhang Technology team are very complementary in terms of perspective and practical experience, which is very important for accelerating the commercialization of Qianhang Technology's autonomous driving technology.

"In terms of technology, Qianhang Technology is still in the stage of discovering problems and solving problems. Whoever is good at what, to solve what, there are differences, our principle is to let the facts, data speak, and the second is who is responsible for who has the final say. ”

The technical framework is being built, the autonomous driving technology should be mass-produced as soon as possible, and the logistics operation also needs to be prepared, which makes Qianhang Technology quite eager for talents.

In addition to the aforementioned Tao Ji, Ding Fei, Sun Haowen, it is revealed that the current Qianhang Technology has also attracted many outstanding talents from different backgrounds, including CVPR best paper authors, NOI gold and silver medals /ICPC world final, kaggle grand master and many other top players, some do deep learning, some do high-performance computing, and some from aerospace/aviation systems and top OEM/Tier1 to join.

Even with the addition of so many outstanding talents, the founding team of Qianhang Technology is still spending a lot of time and energy every day to recruit people, like Liao Ruoxue in order to facilitate the possibility of chatting with candidates at any time, in addition to the need to overcome the jet lag to interview overseas talents, he also specially made a card in a 24-hour tea house near his home.

At the same time, in order to diversify the source of talents, Qianhang Technology also communicates with outstanding talents in the industry and in the non-industry on a large scale, and what is more tempting is that Qianhang Technology is still fully stocked, including fresh graduates and non-product technology positions.

"I will also find friends and colleagues I know before to recommend, everyone's willingness is quite high, at least I have not met anyone who said they were not interested after talking." Liao Ruoxue revealed.

In the once magnificent Internet industry, Liao Ruoxue's whereabouts are like a drop of water, often directly driving colleagues and friends around them, stirring up a circle of water splashes.

Will there be more Internet industry giants involved in the field of autonomous driving next?

Let's do it and see.

END