The Interview – Alexandre Boulenger

Interview and Edited by Jesua Epequin



Alex is the cofounder and CEO of Genify, which provides an API universe for spending analytics and credit decisioning to banks, lenders, and fintechs. It launched its first fintech SaaS software product in 2019 and today has B2B clients on 3 continents, touching thousands of end users weekly. They offer software solutions to banks and Fintechs in two areas: personal finance management and credit decisioning. Their client base is mainly in the Middle East and Europe. Alex has a background in mathematics applied to finance and economics, and a degree in computer science from Tsinghua University. Together, these helped his transition to artificial intelligence. He worked for three years at McKinsey & Company prior to starting Genify.




1. What are the advantages and challenges of founding a startup in China?


Since we are a B2B software company that started just before the pandemic we were not particularly tied to China, and our products were not particularly directed to the local market. That can explain why the company is not in China today.


The core component of Genify is AI, which is central in China. This gets a lot of attention, helps with visibility, and attracts the interest of stakeholders, whether they are VCs, other entrepreneurs, bank clients, and so on.

Moreover, foreigners have access to equity-free government funding in China ($15’000 to $115’000), with no strings attached (besides opening a company locally). To access this funding, you must first pitch your business idea. The pitch event and the grant are different, whoever does well on the pitch goes to the accelerated track to the government grant (foreigners have special tracks, so you don’t compete with locals).  After the pitch, we had to follow a process that took 1.5 years to complete, which in the life span of a startup is a long time. After that time, the money made less of a difference than it would have at the beginning. Although the process can be described as hectic and demanding, the grant is quite generous compared to similar programs in other countries. In short, pitching works but is not easy, it is good not only for the funds but also for visibility: it helps you build a network.


China is very competitive, in other markets you may have 1-10 competitors when you open a company, in China you can have hundreds. For example, a year after Groupon launched in the US, you had around 5’000 companies launching with the same concept in China (Meituan 美团 was one of them). As a foreigner you are welcome, people want you to participate in their events. However, you might get sidelined, you are not expected to add to the critical mass. For example, a bank may invite you to speak at one of their events but may not necessarily consider you as a software provider. Moreover, Chinese banks seem to be less inclined to buy software (they build it themselves instead of buying or borrowing), and in case they do, they pay for the number of people staffed on the project. If you assign 50 people to deploy and integrate the software within a bank’s system, you may get more revenue than if you market smarter software that only requires one person to be deployed, even if it adds the same value.

Finally, it may prove challenging to target other markets while based in China, you may encounter problems related to regulation or perception when selling software to the rest of the world.


2. Does Beijing provide an edge for AI companies?

Beijing is very particular in the AI space because it concentrates all the top science universities, and many tech companies’ headquarters (Baidu 百度, Bytedance字节跳动). In addition, all of them are in a small surface area, within a walking distance from each other in some cases. You have a critical mass of people from the field all in one place. AI also gets significant media and government attention, with most of it focused on Beijing. For instance, for the 2020 annual conference of the Beijing Academy of Artificial Intelligence (北京智源研究院), some of the most famous AI researchers were invited as speakers (among them Geoff Hinton, co-recipient of the Turing prize for his work on deep neural networks). Few other institutes in China can attract such a caliber of scientists. Only in Beijing can you find such an endeavor: a well-funded, highly visible research institute dedicated to moonshot AI projects (such as Wudao 悟道). One may go as far as saying that in China, AI implies Beijing.

The quality of the people I met in Beijing heavily impacted the success of Genify. Belonging to the tech/startup scene during Covid times was proof of guts for both foreigner and Chinese entrepreneurs. Beijing hosts a healthy community around AI, and a dynamic startup ecosystem, from which stems a fertile ground for entrepreneurs in the AI space.


3. What are the key differences between startups in China and abroad?


The first difference is the market size to which you have access. If you target a niche consumer market of 0.1% of the Chinese population, this represents almost 1.5 million people, a huge market by most standards. In other countries, companies often need to target multiple markets, even from day 1, because their home market is too small. For instance, Genify has a fintech client based in Switzerland but starting its business in Germany because the market is ten times bigger there.

Another difference is the shareholder structure to which venture capitalists are used. In the West, in general, VCs like to see the equity split equally between the cofounders. In China, there is a different expectation, it is often the case that the main founder gets more equity, by a large margin.

China is competitive—having a Chinese cofounder is a good idea, however finding one may be difficult. If you are an ambitious Chinese entrepreneur, why seek to work with a foreigner? Your business idea should also be closely related to the Chinese market, and account for the regulatory landscape, customer expectations, and local competitors. As a foreigner, you may have a better shot by first starting a company at home, and then later targeting China. It is not a given that you have better chances in China—rather the opposite, actually.


Genify and AI


4. During your time at BAAI you developed SonoBreast, a software used for cancer detection. How does it differ from current cancer detection methods?


Housed within BAAI, this project was born from a collaboration between Xiehe hospital (PUMCH, 协和医院) and Tsinghua. Its goal is to determine the cancer molecular subtype from ultrasound images of breast tumors. From this, one may infer the most effective treatment and even life expectancy. The molecular subtype is related to gene expression and determines the phenotype (physical characteristics) of the tumor (shape, aspect, and texture as seen in ultrasound images). Today, doctors resort to biopsies to determine the molecular subtype, and cannot determine it from an ultrasound image. A biopsy, besides being prone to errors (because it samples a small part of the tumor), is costly and invasive. SonoBreast aims to help doctors make better and faster diagnosis.

While progressing with SonoBreast, publications on the same subject started to appear in the literature. This put pressure on us but confirmed we were on the right track. The paper we wrote was rejected several  times by different journals and conferences, it was only accepted after 1.5 years: perseverance paid off.


5. You have recently published a paper on bank user segmentation and profiling. Can you explain to us what this is and how it works?


The idea behind profiling is to characterize consumers based on how they use banking services today, and likely will in the future. Banks usually divide users into groups to target them with different products via marketing campaigns. A common way to separate customers is according to age and income level. However, this is far from optimal and may lead to undesirable biases. At Genify, we perform segmentation according to the predicted future purchase behavior, learnt by training a next-best product recommender system. For this, we use embeddings: a numerical representation of the user. The attributes describing each user are encoded in an embedding vector (a continuous, high-dimensional space) in a way that allows us to clearly regroup them according to future purchasing behavior. The product recommendations made by our engine are different between-groups and similar within-group, this shows that the groups inferred via this approach are representative of future purchase behavior. In order visualize the clusters, we generally employ dimension reduction techniques to obtain a two-dimensional representation.


Genify presented this work at the ACM ICAIF ’22 conference in New York in early November 2022.


6. What are the problems Genify is solving?


We solve two main problems via software offered to banks and fintechs. However, those who benefit are the banks’ end users.

  1. If I ask you: how much money did you spend on restaurants, gas, or travels over the past year, could you come up with an answer? Here in China, you probably can if you use WeChat because it keeps track of your expenses. However, in other parts of the world, you may not get the answer unless you spend time (several hours per month) doing your personal accounting. Banks should be the ones doing this, automatically. Transaction enrichment is the foundation to solve this problem: we get data from a bank and enrich it with several fields including a spending/expense category. This allows banks to give their clients a view of their expenses split into different granular categories, but not only this. We can also use this extra information to recommend banking products. Once you know how people spend their money (their spending behavior), you can employ a recommendation engine to predict the most relevant banking products. Our engine makes its recommendations based on the banking products similar users have. In machine learning, this is called collaborative filtering.

    2. When a bank wants to extend a loan to someone, the process is often time-consuming and costly; credit officers at banks often manually review transaction statements. We help banks automate and speed up credit decisions. This is achieved using not only traditional data but also behavioral information gathered from smartphones. This provides banks with robust decision-making, at far greater speed.




7. How did you become involved with French Tech Beijing? What was your role?


I joined French Tech to meet smart and interesting people and expand my network in Beijing. I was involved with French Tech for 6 months. I initially reached out to a former board member, who later invited me to join the French Tech team. My main contribution was (together with another member) to give the impulsion for the revamp of the website. I also joined one of the events as a speaker and helped attract other members. Joining the French Tech helped me better understand the tech scene in Beijing and meet inspiring mentors.

I encourage people interested in joining French Tech to first attend one or two events, get an idea if it is the right community for you, and see how it can complement your network.


8. What do you personally spend most of your time on?


Most of my work time is spent speaking to existing or potential customers, more broadly doing customer discovery—critical for a B2B company. When doing so, I try to speak less than half of the time, to learn from our customers, and get feedback and new ideas, instead of just selling our software products. The rest of my time is spent defining our product and tech roadmap, hand in hand with Genify’s engineering team, building partnerships, hiring, and performing some less gratifying miscellaneous ops and admin duties.


Crédits :  Jesua Epequin (interview and edited), Hugo Menzer & Alex Goncalves digital-space.cn (illustration).