The following interview is a conversation we had with Harry Xu, CEO and Co-Founder of Breeze ML, on our podcast Category Visionaries. You can view the full episode here: $4.6 Million Raised to Create the Future of AI Governance
Harry Xu
Hi, Brett. Hi, everyone.
Brett
We’re super excited to have you here. Harry, to kick things off, can we just start with a quick summary of who you are and a bit more about your background?
Harry Xu
Want? Sure. Of course. I am a professor of computer science at UCLA, and then I’ve been working on education research for more than a decade. So I’ve been working in UCLA. Before that, I was professor at UC Irvine. So basically we’ve been spending a lot of time building systems, computer systems, particularly machine learning systems. So that was one of the reasons why we started the company, to focus on building system related to machine learning.
Brett
What’s it like being a professor at UCLA? What’s something that maybe someone listening in wouldn’t expect? Or what’s the misconception about being a professor?
Harry Xu
Well, that’s a very interesting question. I think life was very interesting because I got to meet and work with a bunch of young people every day because people, like students, graduate, and you always get the new people come in and those are always young people, guaranteed. So I think a nice thing about being a professor is that you always get a chance to work with the youngest generation in the community, and then you have the opportunity to nurture them and turn them into experienced and seasoned researchers and developers, which is actually something that I enjoyed quite a lot. The other thing is that I think a lot of people think that academics are very easy life because you don’t have to produce any product, there’s no sort of monetary aspect to your life because we don’t really talk too much about money.
Harry Xu
So that is actually wrong because the life can be quite challenging because we have a lot of projects. We kind of work as a startup, right? Because you have a bunch of students you have to feed. We also have to talk to national funding agencies like National Science foundation or like Office of the Naval Research to secure funds before you can actually feed the students. So the life is quite challenging, and it’s very similar to how you run a startup. I guess that’s usually the misconception of being academic.
Brett
Now, when you started down this career path, in the back of your head, did you always have this idea that someday I’m going to start a company, or what was the source of the entrepreneurship and the idea to go out and build a company?
Harry Xu
Right. So my Co-Founder, Ravi Naturali, who is a professor of computer science at the Princeton University. So basically, we started the company together. We are both, I would call, sort of atypical academics who care a lot about impact producing impact than producing papers. So if you look at our projects, most of projects do not stop at papers. We always go extra miles and open source our tools and try to get people to use. And both of us had a lot of experience with pushing technology into the actual products for large companies. For example, I worked at Microsoft a few years back, and then I worked on an optimizing compiler that I think is still used in production systems to optimize a lot of jobs on a daily basis. And Robbie has his technology in products at Netflix and Google.
Harry Xu
If you open Google’s Chrome browser, you will see his technology running there. So I think having a company is always something that we’re thinking of. So that’s the reason why we started the company last year, and then we decided that now is the time. Then we have a lot of sort of experience of running those projects, and we have students who graduated last year as well. That’s why we formed a four people team to start our venture at Brazil. Mel.
Brett
And if we talk about timing, there looks like you had great timing. Right? So you launched the company in March 2022, and then the pivotal moment for AI was in, what, November of 2022, when OpenAI released the consumer version.
Harry Xu
Yes, I think, yes. So the chat GBT came out towards the end of last year, and then there was a hype. The hype started sort of coming out early this year, 2023. But back March 2022, the time that we started the company, we actually had a lot of challenges of raising because of the market downturn. There was a market crash in May 2022, and we had a lot of issues in the beginning of raising our seed around.
Brett
Got it makes a lot of sense. I think a lot of founders were in a similar spot there probably as well.
Harry Xu
Yeah. Had we waited a few months, I think we would have had a much better sort of situation.
Brett
Are you surprised with the response to chat, GPT and how it just really seemed to almost take over the world, for lack of a better description. Are you surprised with that impact it had at a consumer level and a business level?
Harry Xu
I’m not surprised, actually. I think it’s one day AI is getting there, but I didn’t really expect that the day comes so quickly. I guess this is sort of. If you ask a lot of people in the community, it’s pretty much what most of the people think about, like the JGBT and generative AI in general. Because one day we all believe that AI is going to become a major thing that takes over a lot of humans tasks, but we just do things back. That the day comes so quickly makes.
Brett
A lot of sense. Just give us a high level overview of what Breeze ML does.
Harry Xu
Yeah, so we’re building AI governance at Breeze ML. So basically the problem we’re trying to solve here is that AI regulations are coming our way, right? So this is something we all know, and the European Union, like EU, has been much more advanced in terms of AI legislation than the rest of the world. The EU AI act is already there, and then they’re looking to finalize the law by the end of this year, and then that’s going to come into effect in the year of 2025. So basically, they give companies two years of time to get themselves prepared for the regulations and compliance in 2025. And the consequence of not being compliant is actually huge. Right? So we’re talking about like a huge fine, something like 6% of your annual global revenue, like uncapped, which is huge for a lot of the companies.
Harry Xu
So the problem a lot of companies are suffering from is they don’t really have any tools, any guardrails that allow them to be compliant as they’re working on the model development. So we’re basically providing this kind of what we call governance by construction tools, right? So developers are using our tools on a daily basis as they are developing models and transformation of the data sets. And our tool provides governance and allows compliance officers and stakeholders of company to quickly gain insight and apply policies over the entire pipeline of your model development and the data transformation.
Brett
Who’s that ICP? Who’s the primary user that really needs AI governance?
Harry Xu
Yeah, I think most of the companies that have anything to do with AI would actually use that because the regulations are generic, right? So there’s some sectors, like some verticals of the industry that are already facing regulations right now. For example, even in the US. I’m not talking about you, even in the US, healthcare is the industry that is facing regulations from FDA. FDA has a lot of policies like gxps, that you have to be compliant with before you can actually release any models in your product. And the other example is the financial industry, like Morgan Leslie’s banks. The banks are facing very strict regulations and compliance from SeC. For example, there are a lot of laws and regulations regarding how your model should be trained, how you should collect user data.
Harry Xu
For example, there’s no way that you can introduce any kind of a bias in your loan prediction model. You cannot really have a model that it was trained over data with general information explicitly considered, for example, because otherwise your model would definitely introduce bias. So things like that are already there. But in general, I think all companies that are using AI are our potential buyers.
Brett
Makes sense. And that’ll be a big market because there’s going to be basically everyone, right?
Harry Xu
Yeah, totally.
Brett
This show is brought to you by Front Lines Media, a podcast production studio that helps B2B founders launch, manage.
Brett
And grow their own podcast. Now, if you’re a Founder, you may be thinking, I don’t have time to host a podcast.
Brett
I’ve got a company to build. Well, that’s exactly what we’ve built our service to do. You show up and host, and we.
Brett
Handle literally everything else.
Brett
To set up a call to discuss launching your own podcast, visit frontlines.io podcast. Now back today’s episode.
Brett
What are you doing to stand out and rise above all the noise? Because this is such a noisy space. There’s just a lot of noise, there’s a lot of funding, and there’s a lot of companies around AI in general. What are you doing to rise above all that noise?
Harry Xu
I think that’s a very challenging question. I think what I do is, I just talked to a lot of people. I had tons of conversations with people doing different things in different roles. Like, I talked to data scientists, I talked to machine learning engineers, I talked to vp of engineering, I talked to compliance officers. I talked to ctos ceos. I think oftentimes once you have a lot of those conversations and they will help you steer your direction of product development. So one of the things, for example, that a lot of people did in the past few months was, if you look at the market, a lot of companies were all in for lms, right? Large language models, because there was a big hype and there are a lot of money in it.
Harry Xu
And I think one thing I believe we did right was, as everybody was all in for lms, we kind of backed out. So we basically started working on this problem of AI governance because of the conversations we had with people. We clearly saw governance is a bigger problem in years to come. And I’m not saying that Loms is not a big problem, it’s also a big problem. But it’s unclear how to build a massive business by building like LM support, for example. So it’s much easier and clearer for us to see a path towards building a massive business around. So I think the key is just to run a lot of conversations and you go with the flow when it.
Brett
Comes to the market category. Is that a category that Gartner has already recognized for AI governance, or is that a category that’s really up and coming and soon to be formed?
Harry Xu
I think AI governance, definitely. I think it’s old kind of existing market AI governance. A lot of people, like I said, everybody that deals with AI have to use a governance tool sooner or later to be compliant with the regulations that’s already there. But the other thing is that I think on the other hand, it’s also emerging market because people don’t know what to do yet, right, because AI governance is a market, but people in this market don’t know what to do at this point. And there are no existing tools. There are very few tools, I would say very few tools out there that can help them provide the governance that need. Because we talked to a lot of lawyers and privacy attorneys, for example, and everybody was talking about auditing AI, auditing models.
Harry Xu
But in terms of concrete steps, the action items, nobody had a good idea of what to audit, what exactly the items to audit. So I think that’s the biggest challenge you have to deal with. You have to be able to concretize the problems first before you talk about how to solve the problem. So I think in some sense is an existing market, but also is an emerging market that people are still trying to define and develop solutions for.
Brett
That makes a lot of sense. Now, can you give us an idea of the type of growth and adoption that you’re seeing today?
Harry Xu
Yeah, so we saw a lot of excitement from our conversations. And we’re doing enterprise sales, first of all. So enterprise sales take much longer than the other type of sales. So we have a few paying customers right now. And then we had a lot of those customers that are trialing our products at this moment. And then what we’re doing right now is that we’re building a sales pipeline. And I hope that once the sales pipeline is built, we can see a very rapid growth in next year or next couple of years.
Brett
As I mentioned in the intro, you’ve raised 4.6 million to date. What have you learned about fundraising throughout this journey?
Harry Xu
Fundraising is hard in general, I think it’s very hard. And then you have to do a lot of things. You have to build connections, you have to assemble a team, your team have to meet a certain criteria before you can raise the fund. Yeah, I think what I learned was I learned a lot, actually. So I think the most important thing is get yourself ready and also raise at the right time. Even if your company doesn’t need money at this point, if the money is there, you just raise. Right? You don’t have to wait until your money exhausts. So, yeah, anytime that the money is.
Brett
There, based on everything you’ve learned so far throughout this journey, let’s imagine you were starting again from scratch. What would be the number one piece of advice that you’d give to yourself?
Harry Xu
That is also a very hard question. So I think maybe I would tell myself to prepare better before seeing investors. Once you see investors, you can’t back out, right? The process gets started, then you have to just follow the process. You do the pitch and you do everything. So I think one thing we could have done definitely better is we can better prepare ourselves. For example, one thing I should definitely do is that if I start the signal one more time, is that I will start paving the road a year before we see the investors, including, for example, assemble the team, clear out the potential IP issues, do the customer survey and then get MVP build and stuff like that. I think a well prepared team has a much higher chance of success in terms of funding.
Brett
And final question for you, Harry, let’s zoom out three to five years into the future. What’s the big picture vision that you’re building here?
Harry Xu
Yeah, so funding wise, we’re aiming to raise our series a next year. So three, five years down the road, I believe that we’ll become a company with several hundred people. I think our goal is to build a massive business. Right. And I’m confident that will be the leading platform in AI governance for both the US and EU market. We definitely want to not only look at the US market, but we also want to sort of explore in the EU market in general. And I believe three to five years down the road will be the leading platform in AI governance for both markets. And most of the large companies will be using our platform. So that is sort of my vision in the next three to five years.
Brett
Amazing. I love the vision, Harry, we are on time, so we’ll have to wrap here before we do. If there’s any founders that are listening in that want to follow along with your journey as you build and execute on this vision, where should they go?
Harry Xu
We do have a LinkedIn, and then these days I’m trying to post more on LinkedIn to be more kind of socially engaged. And then we also have sort of a media blog, and then we start writing medium based blog that we start writing blog articles. But I think LinkedIn is actually the way to go.
Brett
Awesome. Harry, thank you so much for taking the time to chat. Really appreciate it.
Harry Xu
Yeah, thank you very much, Brett.
Brett
All right, keep in touch. Best of luck.
Harry Xu
Yep. Thank you. Take care.
Brett
This episode of Category Visionaries is brought to you by Front Lines Media, Silicon Valley’s leading podcast production studio. If you’re a B2B Founder looking for help launching and growing your own podcast, visit frontlines.io podcast. And for the latest episode, search for category visioners on your podcast platform of choice. Thanks for listening, and we’ll catch you on the next episode.
Twitter
facebook
linkedin