Z Potentials | Exclusive interview with CHAI, CEO of quantitative finance started his second business All in GenAI, with a team of 10 people and an ARR of over 10 million US dollars
In this interview, we are honored to have an in-depth conversation with William Beauchamp, the founder of CHAI AI. A Cambridge graduate who founded a renowned quantitative finance company, William pivoted to the LLM industry after seeing the tremendous potential and opportunities in GPT-2 from OpenAI in 2021. Decisively leaving behind the highly coveted quantitative field, he chose GenAI, asking, “If this technology is so influential, how could it change the world?” Believing that finance could no longer change the world, he turned his focus to AI, an area he was passionate and skilled in, securing funding and computational power from famous capital in his angel round.
In the waves of the Internet and mobile internet, social interaction has undergone revolutionary changes. From the early days of email and instant messaging to the rise of social media, and now to AI-driven personalized communication, technological advancements continue to reshape the ways people connect and communicate. With the emergence of large language models (LLM), we are witnessing the dawn of a new era where AI is not just a tool, but a social partner capable of understanding, generating content, and engaging in unprecedented depth of conversation with humans.
Against this backdrop, William’s CHAI AI product has become a pioneer of change. As an innovative generative AI platform, CHAI allows users to interact with a diverse array of chatbots submitted by developers, covering everything from casual conversations to professional consultations. In this in-depth interview, join William Beauchamp, the founder of CHAI AI, as we explore CHAI’s role in the AI era, the possibilities for future social interactions, and trends in technological development. Let us delve into this conversation together! Enjoy!
01 Seeking a life transformation at age 30 to create something valuable
William: When GPT-2 was released, I interacted with it. Anyone who's interacted with generative AI, whether it's ChatGPT or any LLM, immediately feels it's something special. With GPT-2, I knew right away that this would change how consumers interact with technology. I delved into machine learning and examined the scaling laws. Google had done a tremendous amount of fascinating research and published great papers on scaling laws. OpenAI had also written a bunch of excellent papers. I immersed myself in the literature and quickly gained confidence that this was likely going to be a significant and impactful technology.
I pondered how this technology would change the world and what big businesses and products would need to be built. I felt there had to be a platform for LLMs. Initially, I envisioned it like Google being a platform for websites. Businesses or individuals create a website, optimize it for Google through SEO, and Google matches search queries to the right website. I thought, surely in the future, there would be small specialist teams creating the best LLMs for specific specialties. For example, I might go to the New York Times for recipes or look at Gordon Ramsay's recipes. There needs to be a website that matches these queries to the correct LLM.
I quickly started prototyping, building a small app, and putting it in the app store. The front end was an app where consumers could swipe between different chat AIs or chatbots. The back end, called ChaiPy, allowed users to submit any Python script, with the API being text in and text out. This meant you could leverage existing APIs like Hugging Face and use different models to create your own bot. Each bot corresponded to a Python script. The first ones I made included a news bot and a recipe bot.
Through this process, we quickly saw where the traction was and what the LLMs were uniquely good at. They weren't particularly good at logic or reasoning. Instead, their strength was in being generative, allowing people to interact with them. People loved interacting with them, especially on emotional topics, venting, getting a second opinion, and discussing deep philosophical questions. This shifted my thinking from a platform centered around useful tools to one more like YouTube, optimized for entertainment.
I realized people use platforms like YouTube or TikTok for the feeling they get, whether it's feeling like they're with friends or learning something interesting. Therefore, Chai should be a platform optimized for entertaining conversations with LLMs. As we built in this direction, we gained more traction, reaching around 100,000 daily active users. At that point, I told my wife we needed to move to Silicon Valley to grow the company further.
We got a visa, moved to Silicon Valley, and brought the team with us. At the time, we had around 15 quant traders and machine learners from Cambridge and Oxford. I told them that while they could make good money with trading strategies, this platform for LLMs would be 100 times more impactful. I moved to California and stopped working on trading. A few brave souls followed me, and as ChatGPT gained more prominence, the rest of the team joined Chai. Eventually, only one person remained handling the quant trading.
With this dedicated team, we achieved good velocity and a solid applied understanding of AI. This foundation allowed us to build the Chai platform, gain significant traction, and reach a point where we have a million daily users with only around 10 engineers.
ZP: What motivated you to make this change now, instead of earlier in your career?
William: The way I would phrase this is, I've never really looked at the world through the lens of wanting to build a specific type of product. Instead, I always think about what the biggest, most impactful thing I can build is. When I finished college, I was really good at poker. Poker was about understanding the mathematics of decisions, taking risks, using money, and making good decisions. If you can make good decisions and take appropriate risks in poker, you will make money. I felt I was very good at that and wanted to challenge myself to see if I could succeed. That's why I started trading for myself.
As I invested time into building that business, it went as well as I had hoped. Just as I reached my 30th birthday, I felt I could build something bigger. The question wasn't so much about what I wanted to build or what I was interested in, but where I could have the biggest impact. After working in trading, I learned a lot of Python, programming, and machine learning. In trading, if you can build a good model and trade its predictions, you won't make much money. To make a lot of money, you need the best model. It's a very competitive space.
I devoted a huge amount of time to getting really good and assembling a team that was also highly skilled. We focused on building the best neural networks, random forests, XGBoosts, and feature engineering. With all that experience and the rise of large language models, I felt I could have a big impact using these skills to build something significant. It takes a lot of courage as well. Most people know the right answers but lack the courage to follow those answers to the path they know is right.
ZP: And also, they don't want to take risks. Taking risks is a hard decision for most people.
William: Yes, I think taking risks is about perspective. If you only had one attempt, it would be a big risk. But in life, you can try and fail, but you can try again. As long as you're willing to keep trying and not give up, the risk is quite small. I never viewed it as a big risk because I wasn't relying on one big attempt. I believed something could be built, and I was convinced logically there was potential. I kept trying my best. It might take one attempt, or it might take 100, but eventually, I would figure out something of value. Once you build something of value, the economics will follow.
ZP: Yes, and because you had successful trajectories and positive feedback, you had confidence you could do something better.
William: That's absolutely true. Throughout my career, I started with smaller projects and failed for a long time. Then I took on bigger projects. You start with failure, but through each failure, you learn. It wasn't while the algorithmic trading firm was failing that I decided to do the most ambitious thing. It was once it was successful that I thought, okay, I've demonstrated I can do this. Now I want to apply these skills to have the biggest impact.
ZP: I remember a Chinese saying, "The first brave person enjoys the world." Because you try, you will enjoy the world.
William: Yes, if you think about risk, you can ask yourself what you value. Some people really value security. If I had valued security, I would never have left Cambridge or come here. But if you value growth, having the biggest impact possible, learning, and challenging yourself, then there is no risk. If you do it and fail, you've learned something. I always valued challenging myself, learning, growing, and having as much impact as possible.
02 A user-generated content platform, driven by high quality data
ZP: Okay, let's talk about some questions about the company and also the core product, the Chai application. So, what are you building with Chai? Can you introduce the product features, the number of users, and how they typically spend their time? What are their underlying needs?
William: If you go to the App Store and search for Chai. You'll see an app where you can scroll through a platform of millions of essential bots or Chai AIs, which are created by users. Our most common user is a 20-year-old girl. You might see anime characters, dangerous or scary characters, and people having fictitious, engaging, exciting, and romantic conversations. It's similar to watching a Netflix show like Bridgerton, where users experience drama and various emotions with the AI on Chai.
From a consumer perspective, Chai was the first platform of this kind, and we worked hard to figure out many things that seem obvious now. Unlike companies like OpenAI, Google, and Character AI that aim to train the single best LLM, Chai is more like YouTube compared to Disney Plus. We surface the best LLMs trained by the community. Meta has released fantastic foundation models like LLAMA 1, 2, and 3, and individuals worldwide fine-tune these models for various purposes, such as math, physics, storytelling, and more.
We built a developer platform called Chaiverse, where individuals can submit their models. We have a leaderboard and thousands of LLMs that we evaluate and rank. We serve them to users and get real user feedback, which is invaluable in machine learning. Our goal is to get great LLMs trained by developers, submit them to our platform, and use recommender systems to serve the best models to users. Despite raising only about $10 million, our AI is competitive with those of companies that have spent much more, thanks to our platform approach.
ZP: How do you think we will try to grow in the coming years? One direction I was thinking about is adding multi-modal AI features, like pictures or videos. What are your thoughts on this?
William: When looking at companies and trends, it's helpful to take a long-term view, like a 30-year horizon. Great companies typically take decades to grow, and it's not just about inventing one killer product but about continual innovation and iteration. For Chai, we first allowed users to create their own Chai AI and interact with it through prompt engineering. The second innovation was Chaiverse, where individuals submit their own LLMs, and we use a recommender system to create strong AI.
Looking ahead 30 years, Chai will need to invent and innovate in many ways to grow. Successful companies like Facebook, Meta, and TikTok have continually iterated and innovated. The key is gathering talented people with strong character and fostering a culture of accountability and hard work. Going to work should feel like going to the gym—challenging but ultimately rewarding.
To answer your question specifically, we need to look at successful consumer products like TikTok, Meta, YouTube, and Netflix. Humans love looking at pictures, so AI-generated images must happen. The timing depends on the cost of compute and the sophistication of algorithms. As these improve, there will come a point where incorporating images into the experience makes sense.
Humans also love interacting with faces, so it's likely that Chai will evolve to be more like TikTok, where users talk to it. Text in, text out is great for useful interactions, but seeing a human's face and expressions is more engaging. Audio will also play a role, and it must be social.
ZP: Text may be the fastest way to consume information, but it’s not the only market.
William: I love sending texts because it's easy. There’s no worry about appearance or presentation. But when it comes to content consumption, I enjoy reading blogs, Kindle, and listening to Spotify, podcasts, and audiobooks. I spend even more time watching videos. Human connection is essential, seeing and hearing someone can invoke strong feelings, whether it’s learning, feeling scared, or feeling loved.
In the future, people might prefer interacting with AI-generated faces. While users might be writing text, the responses could resemble interactions on platforms like TikTok or YouTube, rather than WhatsApp.
ZP: I want to learn how you built this application from zero to one million users. One million is a huge number, especially for a team of fewer than 10 people.
William: I'll give you an interesting data point. Yesterday, Sam Altman tweeted about GPT-4.0, mentioning that it processes 50 billion tokens a day. I believe Chai processes about 15 billion tokens a day. So, in terms of raw token processing, we're handling about a third of what GPT-4.0 does. This makes me proud of the number of users choosing Chai daily.
We've never spent money on acquiring users, relying instead on either serving ads or charging our users to fund the company. We’ve received approximately 10 million in external funding but over 25 million from our users, making them our biggest source of funding. Despite having to serve ads, we have significant usage and traffic compared to companies that have raised billions and give away their AI for free.
Achieving such growth is rarely binary. Most things in nature are continuous, and for Chai, going from zero users to 10 was as hard as going from 10 to 100, 100 to 1,000, and so on. Each step of the way requires delivering value. I think about it logarithmically. Reaching a milestone like 1,000 daily active users was tough, and getting to 10,000, 100,000, and more felt equally challenging.
The key is not in finding a single trick that makes everything work but in achieving small milestones. If you can get one user and they find your product amazing, you’ve achieved a milestone. From there, you can aim for 10, then 100, and so forth. This reflects how the world really works—it's not about having a great idea that suddenly explodes in popularity but about gradually increasing the value you offer to your users.
Two and a half years ago, when we had 1,000 daily active users, we focused on understanding the value our product provided and how to expand that value either to our existing users or to more people. We never considered just acquiring users through ads or getting them to share our app. It was always about providing substantial value to individuals.
What has surprised me most is that the strategies advocated by great companies, like being customer-obsessed or building something insanely great, truly work. Ideas based on mere assumptions or cool concepts rarely succeed. The real success comes from understanding what's slowing users down or how to increase the value significantly, and dedicating hard work to achieve that.
ZP: I heard that the initial idea was to build Chai focusing on the China market, using languages like Mandarin or Cantonese. Why did you have this idea?
William: Three years ago, in 2021, I was contemplating the biggest possible market to target. The USA and China are the two largest markets. Typically, there’s a home market advantage: building in America for American consumers, in China for Chinese consumers, and so on.
Building in England would give a home advantage for a smaller economy. Observing trends, it was clear that the Chinese market was growing faster. Thinking long-term, if you build a product in a rapidly growing economy, the business can grow more easily.
My initial plan was to make Chai a China-first product. The app would default to Chinese, and users would switch languages if desired. However, after researching the outcomes of non-Chinese entrepreneurs in China, I realized it might be better to optimize for the American market for greater success.
William: America grows very quickly compared to Europe. California grows even faster. So, California is the richest, fastest-growing economy in the developed world. China is a huge economy with significant potential, still growing rapidly. I think either of these places are strong choices. It was definitely easier to convince my wife to bring the family and the children to California.
ZP: Probably we can talk about Chai's main competitors. Of course, there's Character.AI. Also, there are many startups and applications emerging right now. I would like to know the biggest difference between Chai and those applications. We don't want to say others are not good or bad because every application has its own value. But what's Chai AI's most valuable aspect?
William: Well, I'll start by saying the brutal thing about AI is that companies like OpenAI spend billions building the very best AI, and then they put it behind an API, so anyone can use it for a relatively small price. When looking at the space, you see fantastic app developers or front-end engineers who build a great front end, plug in OpenAI's AI, and immediately have a strong product offering. That's one end of the market.
Then you have companies like Character.AI or Talkie, with large engineering teams who raise a lot of money and train their own AI, optimized for conversations. Chai sits in the middle. We're not so small that we'd be described as just a wrapper, but we're not so large that we can spend hundreds of millions training our own AI. We aim to be a platform.
I often compare it to YouTube versus Netflix or Disney Plus. If someone had told you when the iPhone first came out that video would be a big space and asked which approach—open source, platform, or proprietary—consumers would prefer, you’d see that they love variety. You get big-budget productions from Disney like Avengers, which are professional and widely seen. You get medium-sized offerings from Netflix, which are decent and subscription-based. And then you get platforms like YouTube and TikTok, where content creators around the world produce unique, interesting content that’s more fun.
We've tried to build something fun for creators, whether they are creating chat AI in the app, selecting the bot's image and name, creating a story, or developers submitting their own AI. We aim to create a great creative environment where we can share feedback, give attention, reward creators, and through recommendation systems, provide a unique, interesting, and endlessly fun experience to users.
ZP: So it's more like a UGC platform?
William: Exactly. I think that's an important secret to how a team of only 10 is able to not only compete but continue growing against large, well-funded teams with 100 engineers. It's because we're not the only ones training the AI. We have a thousand models submitted each month on the AI side. We're not the only ones creating the bots. We have millions of bots being created by our users.
ZP: I think one big difference between chat AI and other UGC platforms or chatbox applications is that you can provide user feedback, which can influence product improvement. Can you give me some specific examples of user feedback about this point?
William: Of course. I'll first look at the general principles of platforms and then we can examine how that's applied to Chai. Fundamentally, platforms are always about feedback loops. If we look at a platform like Twitter or X, the feedback loop might be this: you create an interesting and compelling tweet, and recommender systems show that tweet to enough users to measure the engagement rate. If it believes you've created a high-quality tweet, it will give it more reach, and more people will see it. The final part of the feedback loop is that you, the creator, will either receive status or influence. Because this is valuable to you, you'll create more compelling content for Twitter.
An unhappy example is you create a tweet, it fails, the algorithm doesn't promote it, and you get two views—a clear signal that you made a bad tweet. Content creators can learn what's good content and create more of it, or what's bad content and stop creating it.
One of the big challenges we had at Chai was evaluating the large number of language models available—like the 130,000 large language models on HuggingFace. Traditionally, the answer has been to use benchmarks, but we found these don't correlate with user experience. Often, very intelligent models are not funny, interesting, or engaging for users. For example, early versions of ChatGPT were very restrictive, often saying they couldn't answer questions, which was not a delightful experience.
So, we had to experiment. How can we serve these individual models to users and understand whether they are good or bad? There are several mechanisms to choose from. The most straightforward one would be to generate a response from both models you want to compare and simply ask the user which response they prefer. By doing this across our million users and the 100 million messages we generate each day, we can evaluate as many models as needed.
William: Something I want to share is that by seeing user preferences, whether they prefer one model or completion over another, it opens up the door to AI techniques that OpenAI utilizes successfully with their ChatGPT. They took GPT-3.5, did reinforcement learning with human feedback using the PPO algorithm, and that's where ChatGPT came from. At Chai, we share these preferences with developers. They can download the data and use reinforcement learning algorithms on their preference datasets. A popular algorithm is DPO (Direct Preference Optimization), developed at Stanford, which is as effective as PPO (Proximal Policy Optimization) but simpler and easier to use. Developers on Chaiverse are now using reinforcement learning with human feedback techniques, which were previously accessible only to big labs. By open-sourcing the feedback loops and giving them to the community, they're able not only to evaluate their models but also to get great quality data to train their models on.
ZP: For the current model of Chai AI, is it self-developed or using some open-source or closed models?
William: We do not serve a single model on Chai. Any given interaction on the platform could come from any model that a developer created and submitted.
ZP: I think that's a good approach because the models on the market have different features and may not fit Chai. By making small adjustments, they become suitable for Chai AI. We'll talk more about technology. In terms of engineering and implementation, how do we ensure that AI can provide a natural and smooth conversation experience? This is a difficult question for many AI applications. Even with advanced models, it still feels like talking to a robot and not a human. I believe Chai AI makes it feel more human, right?
William: If I understand your question, it's about why interactions with large language models from companies like OpenAI still feel robotic, and how Chai creates an AI that feels more fun and friendly. I see this as a spectrum. I've always found models from other companies to be a bit more fun and to have more personality than OpenAI's models. AI is an incredible tool with many uses. OpenAI wants to be the most useful AI possible, presenting itself as safe and reliable, which is quite different from Chai's goal. When people interact with our AI for fun, factual correctness isn't as important. Some of the most enjoyable conversations are not about being politically or factually correct but about being interesting, funny, or thought-provoking.
Big labs optimize their models for different purposes, not necessarily for fun. Chai doesn't compete with them directly. There are few people serious about AI and machine learning with the objective of making it fun. Many engineers and computer scientists don't focus on the consumer aspect. I'm happy to cater to people who want to use it for fun, whether they're teenage girls, young adults, or lonely individuals. Many engineers don't use popular consumer products like TikTok, so they feel disconnected from them.
I see the value these products provide and am proud to work with a team that builds a platform where people create and share what they love. Chai's AI is different because the company's values are different. We prioritize building things consumers love. If our users love it, we've done a great job. That's different from some other companies.
ZP: Will you consider a Chinese version for the next step to accommodate Chinese users?
William: The really exciting thing about a platform like ours is that there's nothing preventing us from expanding into other languages. If there is a demand for a Chinese-speaking LLM in Chai, someone can train it and submit it to Chaiverse. If users support it, they can score it highly, and our recommender system can then accurately route the correct LLM to the right user.
Although around 5% of our conversations are in foreign languages, only 70% of our users are from English-speaking countries. So there's a gap: 30% of users are foreign language speakers, but only 5% of the conversations are in a foreign language. There's definitely scope for submitting models proficient in different languages. Our goal is to grow the developer community so we can have enough unique language-specific models and refine the recommender system to map the right model to the right users. I think that's a significant step toward growing our platform.
03 "Costco" of AI social platforms, prioritizing value for customers above all else
ZP: Are there any other metrics you can share publicly about how Chai is growing, like three-day retention or other useful statistics?
William: Let me think. What would be good metrics to share?
ZP: Less than 10 people and $12 million in ARR, right?
William: Yes, exactly. My background as a quant trader makes me a big believer in metrics. We A-B test many features and look at their impact. In consumer apps, retention is often the North Star metric, measuring whether people keep coming back.
When we first started three years ago, retention was almost zero. Despite some traction, VCs told us the app was just a fad because people weren't staying. My argument was always that there was demand, but the AI wasn't intelligent enough to retain users. As the AI improved, retention did too.
For instance, at the end of our first year, day 100 retention was around 5%. By the end of the second year, it was 8%, and by the end of the third year, it was 12%. This shows a slow but steady improvement as the AI gets better over time. We aim for long-term growth rather than short-term gains.
We have over 10 million annually in revenue, which is rare in this space, and our unique value proposition has garnered user support. Many subscribe to support Chai even when the ad-supported tier is sufficient. We've also seen consistent user growth; in the last 12 months, while many other products have plateaued, we've grown 2-3x with zero user acquisition costs.
Overall, our steady growth in retention, revenue, and daily active users, despite fierce competition, is driven by continuous AI iteration and refinement.
ZP: Because OpenAI and other cloud services operate on a subscription model, why didn't Chai AI consider this?
William: It's actually about half ads and half subscription.
ZP: Why did you choose to split between subscription and ads, instead of going entirely with a subscription model?
William: If you look at a business I really admire, it would be something like Costco. What does Costco stand for? It aims to give users the lowest possible prices while retaining good quality. For us, it's about making decisions that strike a good balance. Google talks about having a million context-length models, but using such models is unaffordable for what consumers want to pay. We've adopted a "Costco approach," where if we want to double the context length, we have to justify that cost and consider how to pass it on to the users. Regarding the exact form of monetization, it really comes down to each individual. Many people don't have a lot of spare cash and aren't interested in subscribing.
William: But then there are other people with lots of cash who don't mind subscribing. In my personal life, I've been both. When I was a student, I didn't pay for anything and always selected the ad tier. Now, as a dad in my 30s, I just pay to avoid ads. At Chai, we aim to be customer-obsessed. Let's give people the opportunity to subscribe for the best experience without ads. For those not in that position, we'll offer an ad-supported version that remains a profitable, healthy product. Users generally don't mind ads if there's enough value to justify watching a 20-second clip.
ZP: Have you cooperated with TikTokers, vloggers, or Instagram influencers to help promote Chai.ai, or do you focus only on inbound marketing?
William: We do absolutely nothing in terms of outbound marketing. We've always focused 100% on adding as much value as possible. Look at how much advertising ChatGPT did — none. You get the audience you deserve. If you build a really incredible product, you'll get an incredible audience. Worrying too much about marketing or advertising growth means you're not earning that audience. It's far better to put all energy and resources into building a great product. Tesla didn't spend money on advertising for a long time. That's the mindset we hope to emulate at Chai.
ZP: Are you still fundraising for the next round of funding?
William: My view on fundraising and acquisitions is that it's interesting to observe what others are doing, but almost never helpful or productive for us. I've always achieved the best results by focusing on what we're doing and thinking clearly about the best course of action. When I first arrived in California, I spent a lot of time talking to potential investors, which distracted me from focusing on customers and their problems. I realized that the cash offered wouldn't change the product or make users happy. The true funders of our company are our users. We're not VC-driven; we're customer-driven. Obsessing over delivering value to customers has consistently led to great outcomes.
William: A lot of startups raise a lot of money but eventually focus too much on VCs instead of their products and users, which isn't good. From a common-sense perspective, a business should strive to be profitable or at least have good unit economics. If you're spending $10,000 a day on GPUs, you should have $10,000 a day in revenue from users. Chai is profitable and has solid unit economics. There’s no pressing need for funding at the moment. However, as our hiring increases, especially for talented engineers from companies like Meta, there may come a time when additional funding is necessary. Our balance sheet is healthy, allowing us to attract a lot of talent. Funding should first come from your users. If you're growing and need more capital to attract top talent, then it makes sense to speak to VCs, but never the other way around.
ZP: So what kind of talents are you currently looking for?
William: We're primarily looking for people with experience in consumer products, such as engineers from Meta and TikTok. We recently hired an E7 from Meta, a very senior and incredible engineer.
William: And it's not only that. If you think of it as saying, we have 1,000 straightforward tickets to do, so we can just hire anyone. That's really not the case. We need really smart, creative, talented people who can build foundations, build feedback loops, iterate, and understand problems very well. I find that if you have a very talent-dense team and add people with less talent, you'll actually slow things down. So, aiming for talented people, typically from a consumer background, who live within 15-20 minutes of Palo Alto seems to be the recipe for success.
ZP: Are you still looking for some AI researcher or top research PhD, like this kind of talent? Or are you still looking for engineering talent?
William: That's a fantastic question. We have a consumer app, so if you've built an app, there could be a great fit. A lot of people ask about having a website. If you've built a website, there could be a great fit. There's a lot on the back end, such as figuring out feedback loops. If you've built a recommender system, or done traditional machine learning, there could be a great fit. If you've worked on ads or monetization, there could be a great fit. With AI, there are important questions, whether it's safety, recommendation systems, or techniques we can share with the community. If you're a talented engineer, we can almost surely use you productively.
William: Companies without expertise in AI and machine learning feel they need to hire PhDs. I've spent a decade in machine learning and AI in Quant Trading, which is something we're very good at. I found PhDs aren't necessarily the best fit. I’ve never built an app before this, and we have less experience in app and web development. AI is a strong point for us.
William: Yes, I think a lot of people think it's an AI problem when it's much more of an engineering problem.
CHAI AI is currently actively recruiting passionate AI professionals in Palo Alto to jointly develop the next trend-setting product. Our team needs the following professional talents: Senior/Staff Software Engineer (Flutter/ML Infra/AI/Fullstack/Backend). If you are interested, please send your resume to shiya@chai-research.com.
Z Potentials will continue to provide more interviews with entrepreneurs in the fields of artificial intelligence, global markets, robotics technology, and more. We invite you to join our community, share, learn, and grow with us.