Co-founder and CEO of Liulishuo. He left a Google product role in Mountain View, flew home to Shanghai, and shipped an AI English teacher that ended up on tens of millions of phones.
Open the Liulishuo app and a voice asks you to read a sentence aloud. The app scores you on rhythm, stress and vowel length within seconds. The judge is not a teacher in a chalk-dusted classroom in Hangzhou. It is a model trained on millions of recordings, the latest of which might be the one you just submitted thirty seconds ago. The man who pointed the company at that strange and patient idea is Yi Wang, and he has been pointing at it since 2012.
Wang is the co-founder and chief executive of Liulishuo, known internationally as LingoChamp and traded in New York as LAIX. The company sits in Shanghai, on Changyang Road. The product sits in pockets. The promise sits in the brand name: Liulishuo is Mandarin for "speak fluently." Wang built the entire stack around making good on that promise without hiring an army of human teachers to do it.
His CV looks engineered for the job. A bachelor's and master's from Tsinghua University in electronic engineering. A 2009 PhD in computer science from Princeton, where he worked on the kind of networking problems that are invisible until they break. Two years as a product manager at Google in Mountain View, where he helped shape early Google Analytics features, including the dashboards and widgets that quietly trained a generation of marketers to think in funnels. He holds a US patent on virtual router migration. None of that screams "language app," and that is exactly the point.
In 2011 Wang did what the Chinese press calls becoming a "sea turtle": an overseas student who swims back. He returned to China with a hypothesis. Conventional English schools in China were expensive, theatrical and inconsistent. Smartphones were everywhere. Speech-recognition models were getting good enough to grade pronunciation, not just transcribe it. The natural product, to Wang, was not a school. It was a tutor that fit in your hand and listened all day.
He launched Liulishuo in September 2012 with two co-founders, Hui Lin and Ben Hu. The first version of the app shipped in 2013 and climbed to the top of Apple's China App Store within months of release. By 2017 the company had pulled in around 100 million dollars in funding, crossed roughly 50 million registered users, and Princeton was running press about the Tigers alum who had quietly built one of the country's most-downloaded education apps.
A year later, on September 27, 2018, Wang stood on the floor of the New York Stock Exchange. LAIX Inc. priced its IPO and raised about 71.9 million dollars on a market cap near 600 million. The ticker on the board read LAIX. The bell, by any sensible measure, could probably have been graded by his own product.
The thesis behind the company is unfussy and a little radical. Wang has said publicly that Liulishuo's AI teacher can roughly triple a student's learning efficiency compared with a traditional human-only setup. He is not arguing teachers are bad. He is arguing that the bottleneck in language learning is not lesson content; it is supervised practice. A teacher cannot listen to one student for six straight hours. A model can. A teacher needs a salary. A model needs cloud time and a sane loss function.
That is also why the company has accumulated keywords that read like a Silicon Valley menu: speech recognition, big data, app monetization, SDK and API integration, B2C, D2C, B2B, gamification, SaaS. Behind the buzzword cloud is one operating belief Wang carries from his Google years: the right user experience is a feedback loop that closes faster than your attention drifts. Liulishuo's product is, fundamentally, a feedback loop disguised as a tutor.
Liulishuo growth, by reported milestone
Wang left Mountain View in 2011, when AI consumer products were still rare in China. Friends called the move early. The App Store charts of 2013 called it correct.
"Liulishuo" literally means "speak fluently." The brand makes a promise before the splash screen finishes loading.
Wang holds a patent on virtual router migration from his Google era. The instinct for moving traffic without breaking the link now powers a teacher that never drops the call.
Hui Lin and Ben Hu co-founded the company with Wang in September 2012. They have been the trio of record ever since.
Wang took the Liulishuo story to the 2018 Stanford China Education Forum, framing the product as a "new era" of teaching - one in which the loudest part of the classroom is the student.
Shanghai's IT establishment names him to its annual list of young leaders. Wang stays largely out of the gossip press.
"Redefine language learning through technology and gamification."
"You are unique. You have a gift. Our job is to help you find that and realize your full potential."
"In China you have one standard - get the best grades and you will have a successful life. But in the U.S., everyone seems authentic to themselves."
Read the three quotes back to back and you get a working sketch of Wang's operating style. He treats product as a teaching problem. He treats hiring as a coaching problem. He treats culture as the gap between Princeton seminars and Tsinghua exam halls, and he assumes the right startup sits somewhere in between. Liulishuo is the prototype.
English is the country's second economy of attention after social. Tens of millions of students, parents and professionals practice it daily. The conventional answer is offline schools, after-school programs, tutoring chains. Wang's answer is a phone that listens. That move sounds obvious now. It was not obvious in 2012, when speech models routinely mis-heard accents and the consumer AI market in China barely existed.
The financials around Liulishuo have not always been a clean fairy tale. The company has navigated the broader weather of Chinese education tech, including the 2021 regulations that reshaped after-school markets. But the technical bet aged well. Speech models got better. Phones got faster. A generation of learners grew up expecting the teacher to be in the app, not in the room.
Wang's company sits in the small group of Chinese AI startups that shipped a real consumer product before "AI consumer product" became a phrase a venture firm could spell. The Liulishuo team has been adding features around adaptive lesson plans, AI-driven practice and SDKs that other apps can plug into. The phrase Wang used at Stanford - "new era of learning" - is now closer to a description than a slogan.
What Yi Wang is doing, in plain language, is operating one of the longest-running consumer AI bets in Chinese tech. Speech recognition has been the company's anchor since day one. Generative AI is increasingly being layered on top: better dialogue practice, better explanations, more patient correction. The product still does the same thing it did in 2013 - it listens, judges, encourages, repeats - but it does it with more depth each year.
There is also the matter of the founders staying in place. Many Chinese consumer apps churn through CEOs as their boards search for the next growth story. Wang has been at the helm of Liulishuo since 2012. He has not pivoted into self-driving cars, crypto exchanges or live commerce. He has stayed in the same lane. That kind of stubbornness is rare and, if you spend any time with the company's product, visible in the details.