How an AI entrepreneur deals with dirty real-world data

Women in the AI field are making research breakthroughs, spearheading vital ethical discussions, and inspiring the next generation of AI professionals. We created the VentureBeat Women in AI Awards to emphasize the importance of their voices, work, and experience, and to shine a light on some of these leaders. In this series, publishing Fridays, we’re diving deeper into conversations with this year’s winners, who we honored recently at Transform 2021. 

Briana Brownell, winner of VentureBeat’s Women in AI entrepreneur award, didn’t set out in this field to earn accolades. She set out to create an AI that would do her job for her — or at least that’s the joke she likes to tell.

Really, she set out to build a company that would combine her data analytics background with AI. In 2015, she launched Pure Strategy, which uses ANIE (an Automated Neural Intelligence Engine) to help companies understand unstructured data. She and her team invented algorithms from scratch to make it happen, and the system has been used by doctors to communicate with patients and with each other across cultural knowledge, for example. She also moonlights as a science communicator, inspiring not just young children — especially girls — but everyone around her.

“Whether you’re interested in the intricacies of algorithms to validate unsupervised machine learning models or a high level future view of humanity and AI, Briana makes you feel comfortable with her genius,” said Roger Sanford, CEO of Hcare, who nominated her for the award.

Brownell told VentureBeat she’s “extremely excited to have won this award.” “It’s a huge honor to me,” she said. “It was definitely a surprise because I think the competition was pretty fierce.” Indeed it was, but we’re pleased to recognize Brownell’s work as an AI entrepreneur, and even more excited to further chat with her about her work, the role of AI entrepreneurship in the broader field, and bringing more women to the table.

This interview has been edited for brevity and clarity.

VentureBeat: Tell us a little about your work and approach to AI. How did you come to launch Pure Strategy? And what drives you overall?

Brownell: ​​I started Pure Strategy after spending about 10 years as a data scientist. I was still doing a lot by hand, but there were new techniques coming out that made working with some of those data sets a lot easier. You started seeing natural language understanding, and neural network infrastructure became available in open source packages. All of that really just accelerated. I jokingly said I wanted to essentially program myself into the computer so that I could create an AI that would do my job for me. And that’s essentially what I set out to do — try to use those technology tools to make it easier and faster to do data analysis.

VentureBeat: And when you were creating your product ANIE, what were some of the challenges you faced? And how did you overcome them?

Brownell: There were a lot of challenges for sure. The first was that many of the algorithms we use weren’t actually invented yet. And so we have a whole suite of proprietary methods that make our platform perform at the level it needs to. And so that was really a challenge because it was a lot of trial and error and a lot of building the system out so that it would generalize to a lot of different cases. The second one was being able to find and analyze the data that we needed. The size and scale of the datasets we use for training made it extremely difficult to program things efficiently. I would, let’s say, set a neural network to train, and then I’d have to wait 20 or 30 minutes for it to do the first step. And so that took a lot of time and was a real challenge.

VentureBeat: How do you view AI entrepreneurship versus academic AI research and other aspects of the field? What are their unique roles, and how can they best come together?

Brownell: I think one of the challenges people have in going from AI academia to entrepreneurship is that they are very, very good when the data is all correct, the algorithm fits the assumptions of the modeling, and everything is sort of beautifully positioned to fit the case. But in the real world, everything is incomplete and data is dirty. You may not be able to find the data that you need, or you might have to find a way to approximate it. You might have to merge data sources. All kinds of little issues come up when you’re working with real data, and that’s where I think my experience working in the industry, with lots of different kinds of data, and lots of different kinds of problems with data, really came in handy. Because when you’re building a platform that you’re going to try to get a company to use, it doesn’t matter if it’s the perfect algorithm academically; it matters whether or not it works and if it helps the company make the right decision. And so I find that it’s increasingly difficult for people to be really strong in both business outcomes and the theoretical AI area. And so we need translators, essentially, that can work across those lines and understand what’s possible with AI and what’s relevant for the business. So that intersection is really, really important.

VentureBeat: Do you have any pieces of advice for AI-focused entrepreneurs. What often gets overlooked? Or what’s something you wish you knew earlier on?

Brownell: It’s easy to create a general model that will do something, but it’s very difficult to customize that model to work in a specific case and do that at scale. If you look at all major AI company failures, and I don’t know if you’ve followed Element AI, for example. But they had [$257 million] in funding and all this amazing talent, and they struggled with that. And I think that we all underestimate how valuable that customization actually is. I think that’s a critical, critical factor. Big companies really struggle to get their heads around AI because there’s no guarantee it’s going to work. They love to make these huge claims to get in the door, and then so many of these projects fail because they’re over promising. And so I see that as a big threat to the industry. The graveyard is littered with AI companies that have made huge claims.

VentureBeat: Your nominator said you’re often the only woman in the room, which is of course common for women in AI and in tech more broadly. There’s long been talk about this problem and the risks when it comes to AI in particular. But do you feel like anything’s changing? And how does it all play into these ongoing discussions around the importance of ethical and responsible AI?

Brownell: At my first job, which was in finance, I was the only woman who worked at the whole company, actually. And at my next job, I actually worked for a female CEO with a lot of women technical staff. And so I thought women in data science and analytics was just the normal state of the world. And then I got a rude awakening when I got into tech. And I think it’s a real shame because there’s a lot of promise with how AI can change societies and the world. And not just more women, but people from underrepresented groups overall at the table, can help us solve problems that can’t just be solved when you have group think. And so I’m hoping that as more women start becoming prominent in AI, the types of use cases start becoming more interesting and that more women choose this career. Because there’s a huge need for diverse perspectives and new ways of thinking about how the technology impacts our lives.

VentureBeat: You’re also working on a children’s show that revolves around explaining complex science topics — like AI — to pre-teens. How did you get into that, and why is science communication important to you?

Brownell: It’s extremely important to me. I actually have a few other things I’m working on in that area: I write about physics and astronomy for Discovery, develop K-12 AI content with charities to make it more fun and accessible, and am working with TED on AI explainer videos for kids, too. I think reaching students when they’re young is really important, because you don’t really know what careers are possible when you’re growing up unless you see it in your inner circle. I worked with an engineering association called APEGS, which has a program to encourage more women to consider engineering. And one of the things that they talk about is that a lot of the women who decided to go into engineering, they had a relative or close family friend in the field who could see their skills and encourage them. And so being able to expose people to the kinds of careers that are available, I think, is really critical.