Interview with Former Head of Kimi AI Search: Targeting AI dating and marriage startup, secured tens of millions of investment from the Queen of Venture Capital within 3 hours

According to a report by Zhidx on July 28, recently, Shenzhen-based "AI + dating" startup Liangpei Technology announced the launch of its product "Liangpei". Previously, it obtained an angel round financing of 15 million yuan with the investor being Capital Today.

Zeng Xinxun, founder and CEO of Liangpei Technology, told Zhidx that he initially planned to raise 10 million yuan by offering 15% of the shares. After about 3 hours of communication with Xu Xin, founder of Capital Today, Xu Xin proposed to acquire 20% of the shares and increased the investment amount to the final 15 million yuan.

The completion of this round of financing was quite coincidental. Before that, Zeng Xinxun had contacted about 10 investors, and the feedback he received was not optimistic. Some people thought the scale of the dating and marriage market was limited, some questioned whether emotional problems could be solved by technology, and others worried that a pure technical team lacked the ability of market promotion.

During a hackathon event in Hangzhou, Zeng Xinxun ran into Evan, an investment manager of Capital Today, in a coffee shop. The two only talked for about 15 minutes, and the next day, he got the chance to meet Xu Xin.

Zeng Xinxun recalled that Xu Xin had previously invested in BOSS Zhipin, and her husband Li Song is also the founder of Zhenai.com, who has long-term observation on the two "people-finding" businesses of "job hunting" and "dating for marriage". After the rise of the AI wave, Capital Today has been looking for teams that can use AI technology to transform the recruitment and dating & marriage industries.

"When I walked down from the upstairs, I felt that financing and blind date are exactly the same, which requires mutual effort from both sides." he said.

Liangpei Technology was founded in 2025, headquartered in Shenzhen, and is positioned as an AI Native product for people who are looking for serious relationships. Its first mini-program version has entered public beta, and the App is under review in the app store. As of the interview, Liangpei has been online for about 10 days, with more than 9,000 users.

Interface of Liangpei mini-program

Compared with common dating platforms, Liangpei has set several quite radical rules:

Users need to have a 20-minute communication with the AI matchmaker during registration;

Each user can only establish a matching relationship with one other user at the same time;

Members can choose the payment plan of "full refund if no marriage".

The company also tries to train a model with quite specific goal: when inputting the profiles of two users respectively, the model will output whether the two are suitable for each other and the probability that the relationship may succeed.

Zeng Xinxun previously was responsible for search and recommendation business at WeChat and TikTok, and his last job was technical lead of AI search at Kimi. In his opinion, looking for a partner is essentially a search problem between people. Traditional dating platforms can expand users' contact scope, but lack the ability to understand complex personal information and relationship needs.

When he left Kimi, the value of his unvested stock options was about millions of yuan. With the rise of Kimi's valuation, the corresponding value of these options has now reached tens of millions of yuan.

He did not show much hesitation about this choice.

Back when he was a sophomore at Southern University of Science and Technology, Zeng Xinxun suspended his studies to run a food delivery platform. The project covered Shenzhen, Dongguan and Huizhou, accumulating tens of thousands of users, but later failed due to large-scale subsidies from Meituan and Ele.me. After that, he returned to school to continue his studies and transferred to the computer major, then successively joined WeChat, TikTok and Kimi.

After the failure of his first startup, he has been waiting for the next opportunity. Until the stability, usage cost and supply ecology of large models changed, he believed that the conditions required for AI application entrepreneurship were ready.

The more direct product inspiration comes from his own more than one year of online dating experience.

Zeng Xinxun once wrote a 2000+ word personal introduction on a dating app, but still experienced a lot of inefficient communication. Just when he was about to uninstall the app, he met his current wife. Both of their personal profiles were written in great detail, they decided to meet after chatting online for one day, and confirmed their relationship on the first meeting.

Group photo of Zeng Xinxun's family of three (1st birthday commemoration of his daughter)

This experience made him form a judgment: What online dating lacks is not more candidates, but more sufficient context, deeper platform understanding, and a more focused communication environment.

Focusing on these problems, Liangpei has designed functions such as AI Matchmaker, AI Avatar, AI Advisor and AI Recommendation, and set "average user success rate of finding a partner" as the North Star metric of the product.

Zhidx had an in-depth exchange with Zeng Xinxun for nearly two hours. We tried to understand why a search technical lead chose the seemingly traditional industry of serious dating; how large models process a large number of vague, implicit information that even users themselves do not realize in mate selection needs; and why a dating platform actively limits user activity and promises to refund membership fees if no marriage is achieved within three years.

The following is the sorted interview transcript on the premise of not changing the original meaning, provided by Zhidx.

01. Xu Xin invested 15 million yuan in 3 hours, the team took a call and went to Shenzhen

Zhidx: You met Xu Xin for the first time and talked for three hours, and quickly finalized the investment. What happened at that time?

Zeng Xinxun: Before meeting Xu Xin, I met about 10 investors and received various objections.

Some people thought the market was not big enough; some thought blind date is more about emotion and hard to be solved by technology; some thought AI is just a conceptual packaging; others thought the dating and marriage industry relies on promotion and growth, and I have a technical background, so I might be able to make a product but fail to sell it.

After meeting Xu Xin, none of these doubts appeared.

I hoped to find an investor who truly recognizes this industry, and she was also looking for people who have the ability and willingness to do this thing.

Capital Today previously invested in BOSS Zhipin, and Xu Xin's husband is also the founder of Zhenai.com. She has long-term cognition of both "job hunting" and "dating for marriage". After the arrival of the AI era, she has been looking for teams that use AI technology to transform the recruitment and dating & marriage industries.

My experience perfectly fits this direction. I have done search and recommendation, and was responsible for AI search. I have also experienced the setbacks of online dating, and finally met my wife through a dating app . I have experienced grassroots startups, large internet companies, and AI unicorns, with more diverse experiences than most people.

So it's hard to say who impressed whom. She was looking for such a team, and I was also looking for an investor who understands this matter, and we hit it off immediately.

Zhidx: How was the amount of 15 million yuan determined?

Zeng Xinxun: My initial financing target was 10 million yuan, offering 10% of shares, corresponding to a valuation of 100 million yuan. After talking for three hours, she asked if she could take 20% of the shares.

My first reaction was whether she wanted to take 20% of the shares with 10 million yuan. At that time, I thought that the opportunity to get investment from Xu Xin was very rare, and it was acceptable to give an extra 5% of shares, so I agreed quickly.

Then she took out a piece of paper and started to list formulas. I realized that she planned to increase the investment amount according to the original valuation, and had no intention of taking 5% of shares for free.

The final calculated amount was 14.16 million yuan. She said to round it up, and finally invested 15 million yuan, accounting for 20% of shares, corresponding to a post-investment valuation of 75 million yuan.

Later, when I contacted other investors, I found that there are very few cases where investors actively raise the price for entrepreneurs, and most people will only suppress the price.

Zhidx: How was the core team of Liangpei formed?

Zeng Xinxun: The initial team had 7 people, and now there are 17 people.

Group photo of Liangpei Technology founding team (7 people in the initial team, 5 of whom resigned and moved to Shenzhen from other cities)

After the failure of my first startup, during my years working in large companies, I have been consciously getting to know people who may start a business together in the future. Everyone knows my work ability, character and way of doing things, and also knows that I will start a business sooner or later.

This time I got the financing about a week after I decided to start a business, and then began to recruit the team. For these colleagues and friends I have known for many years, this is not a sudden invitation. They have observed me for many years and are psychologically prepared for starting a business.

Most of the people in the team decided to join after I introduced the project to them on the first call, and only one member communicated more than twice.

Our operation lead Wang Qian has been starting businesses continuously for 7 years during college, and once worked at Procter & Gamble, good at quickly finding solutions from ambiguous problems. Many things in the team without clear methods will be explored by him first.

The technical lead is my junior fellow apprentice at Southern University of Science and Technology. We published papers and participated in competitions together when we were in school, and long-term cooperation built trust. He is mainly responsible for engineering research and development.

I used to do technology and algorithms, but I have always had strong product thinking, so now I mainly serve as the product lead. What functions the product has and why it is designed in this way are mainly my responsibility.

The growth lead participated in the process of Soul growing from 0 to 100 million users. We talked for only one hour, and both sides decided to cooperate on the spot. About half a month later, he moved to Shenzhen from Beijing.

Our current goal is to build the leading serious dating platform in China first. When we supplement overseas talents in the future, we also hope to bring the product to more markets.

02. Resigned from Kimi to start a business, the opportunity I have waited for more than ten years has come

Zhidx: Your experience spans a wide range. You suspended your studies from Southern University of Science and Technology to start a business when you were a sophomore, then worked at WeChat and TikTok in charge of search-related work, and finally served as technical lead of Kimi AI search. Why did you give up Kimi's options and cross into the dating and marriage industry?

Zeng Xinxun: When I left Kimi, that part of the options had not been worth tens of millions of yuan, and it was worth about millions of yuan at that time. Kimi's valuation has increased over the past year, and it is worth tens of millions of yuan according to the current valuation.

My first startup was in 2013, when I was a sophomore, and I ran a food delivery platform. Later, Meituan and Ele.me entered the market and started large-scale money-burning subsidies, and our project failed.

Zeng Xinxun delivering food by himself on a typhoon day (left) and orders during peak hours (right)

After the end of my first startup, I already knew that I would start a business again in the future. I returned to school to continue my studies and re-learn computer science, which was also to accumulate abilities and look for opportunities for the next startup.

However, when I graduated, the mobile internet was almost mature. Products of this generation such as Pinduoduo, Xiaohongshu, Kuaishou and Douyin had already appeared, and it was difficult to find new entrepreneurial opportunities in the market. So I could only enter large companies first, accumulate experience, resources and contacts while waiting for new opportunities.

No matter whether I went to WeChat, TikTok or Kimi, I told my leaders and HR in advance that I would start a business in the future. I will give full play to my value in the company, and I will leave at the right time.

WeChat search team building

After ChatGPT came out at the end of 2022, I have been looking for application entrepreneurial opportunities in the AI era. By 2025, I thought the opportunity suitable for me had appeared. No matter how many options I have or which company's options they are, it will not affect this decision.

Zhidx: Why do you think 2025 is the right time for AI application entrepreneurship?

Zeng Xinxun: When ChatGPT first came out in 2023, I started to pay attention to AI application entrepreneurship. My background is partial to application, not training foundation models, so the opportunities are mainly at the application layer.

I participated in one of the earlier domestic AI hackathons that year, with more than 300 teams on site, and we entered the top ten. After the competition, I found that there were three problems with the large models at that time.