*APAC Heritage Month* This is Your Tech Leader Speaking

Hi there, I'm Gretchen. Thanks for joining us for today's session of This is your Technologist Speaking. This is a series of talks with local APAC industry leaders. We've been lucky enough to curate a guest list that is filled with technologists both with knowledge and leadership experience. Today's guest has leadership and strategy experience in artificial intelligence, machine learning, and data science. She is an astrophysicist and a mathematician who currently heads up data and AI at Whisper. I've got to say, I never know whether to say data or data with my accent. Previously, Fiona was with ELISA, which is one of Australia's most prestigious machine learning and AI organizations run by truly inspiring people. Fiona is also the founder of Women in Machine Learning and Data Science, whose mission is to support, promote, and develop women and gender minorities practicing and working in the field. Fiona, you have an amazing bio. Every time I look at it or hear you coming to talk, I'm like, oh my gosh, you've achieved so much. It's brilliant. But let's dive right in. You're the head of AI and data at Whisper. What does Whisper do? And I guess, what do you do at Whisper? Yeah, sure. So Whisper is a B2B communications platform. So businesses use us to reach their audiences. There's a wide range of use cases. So yeah, it kind of ranges from something pretty boring, like one-time passwords, right through to more interesting stuff, like telling people when to leave the area when there might be bushfires around. And the company was founded 20 years ago by an architect. They've kind of been quietly achieving since then. So what Jeremy, so the founder, who is now also the CEO still, he's an architect. And he was having trouble kind of communicating with all the stakeholders in his projects. So he decided that he needed some kind of software to help solve that problem. So he just built it, which is a pretty cool story. But yeah, at Whisper, we've got vast amounts of text data and engagement data. So we can kind of tell you what people are doing when they receive their messages. And yeah, so there's a big recognition that data science can be applied to really help our customers kind of write the right message. But not only that, send it to the right people at the right time, via the right channel, so that engagement's up. But not only is engagement up, but we're encouraging that our customers not to be spamming their audiences, which is something that... Such a good point. Yeah, it kind of helps me sleep at night knowing that I'm encouraging people to send less messages, which is cool. But yeah, we've got... Sorry, you go. I was just going to say that. So there's currently a team of five, soon to be seven. And we are looking for one more person as well. But within the team, we really kind of, we've set up to marry up engineering with science. Very few people are really good at both. And both kind of skill sets really do well when you put them together and let them collaborate and let them learn from each other. It's something that I learned working at Eliza. And it's definitely something that I'm trying to foster at Whisper. Yeah, because they are two different skill sets, but you need a little bit of both to actually get the right results at the end. I'm intrigued by the bushfire text messaging you mentioned. So are you saying you can take data from that? And then if this happens again, the messaging will be more timely and actionable? Is that the end result you're hoping for? I mean, so the bushfire use case, it's the fire service that use us. So yeah, we can certainly help. We can be looking at helping them with location data and making sure that they can, I suppose, organize their contacts in a good way so that when they're ready to pull the trigger and tell people to leave, that we've given them the tool to allow them to do that. But that's not to say that we are kind of fire experts. You wouldn't want to be pulling the trigger too early on that one. So that might not be the perfect use case for helping our customers kind of send less messages. That kind of use case is maybe more along the marketing end of the scale. So it sounds like you do, I guess, the maths, ML, AI side of things at work. But actually, also, there's a real practical application, which is that to your point of needing a bit of a diverse team there, not just engineers and just machine learning people. Yeah, so I mean, I'm quite a practical worker in general, or a practical person in general. And in data science, it is pretty easy to kind of, if you want to get caught up in the academic side of things, or the theoretical side of things. But yeah, I'm just kind of, I'm a practical, pragmatic person. So I'm always looking at how can we apply this thing to real life? Yeah, exactly. I love that. Hey, what do you think the biggest challenge in data science is at the moment? Yeah, that's a good question. So probably the same challenge that we're seeing across tech, and that's just ethics, really doing the right thing or trying to do the right things. So yeah, I think that there's so many reasons why this is a challenge within the industry. But ethics is hard. You just have to kind of look at the trolley problem where you have to decide to save one person that was going to die anyway, to save two people. And then, yeah, it's really hard to a problem to solve. And then the types of systems that we're building within tech are really powerful, because they're scaling so much, and they're potentially making decisions for people. And so you just need to like, mess up a person, a point a tiny bit, and you're like screwing over hundreds and thousands of people potentially. I mean, we saw what happened in America with elections. Yes, that's a, an extreme case. But yeah, these systems are really powerful that we're building. And sometimes the implications aren't always completely thought through. And sometimes it's impossible to know what the implications are until they're out in the wild. But unless you've got an ethics lens in your design process, you're definitely going to potentially build something that could cause harm, or if it gets into the wrong hands, then that person can use it for harm. And there's another part of the problem is that there's definitely not enough diversity at the table. And so you can't, you can't, if you, if you grow up privileged, you can't always think about the negative impacts that a system that might perpetuate problems, people coming from different backgrounds, if they're not sat at the table to talk about those implications, or recognize that those implications might be had, then again, mistakes will be made and people will be disadvantaged because of that. And yeah, I think the government as well as in technology, it's the government's generally playing catch up. So yeah, I think we're getting a lot better in recent years in the government. But yeah, I think they really need to get completely caught up so they can have more of a say in what, especially the bigger tech companies are doing. My experience in that space is that the governments just don't have a true understanding of what's happening. They don't have the people and the knowledge to advise them as they need. And it's, yeah, it's catch up, it's catch up time. Hey, how do you think it's possible? And this is a bit off topic, actually, to hold space for those conversations around ethics, especially if you've got a diverse team without people going, wow, this is wild and crazy and offensive, and I'm just going to stop doing it and build my product because that's easier. Is there a way? Yeah, I mean, there's probably lots of ways you can, I mean, I can talk to how I'm trying to do that. And that's, yeah, baking it in from the beginning. So I'm in quite a unique position where I've been able to build my team up from scratch. So I'm not really inheriting a lot of like, bad behavior or bad habits or ways of working that I need to kind of change. So yeah, I'm kind of starting from the ground up. So bake it into the design process, make sure that the team is aware of the problems that they can come across with the things that they're building. Yeah, it can come down to who you're hiring. So in the interview process, you can ask questions around that person's knowledge around ethics and make sure that you're hiring people that understand the importance of ethics. Yeah, I suppose it's about working at companies that will allow you to foster that culture. We're kind of in a good position at the moment because ethics is like trending and it's cool and buzzy to say that you're thinking about ethics. So a lot of companies will try and encourage you to think about that. But if that means you telling upper management that you can't build something because you don't believe it's ethical, that's when the real kind of put your money where your mouth is situation is. And at Whisper, I haven't really had that situation yet. But yeah, I think that's something that it would be the hardest thing to do. But yeah, I suppose it's having that process in place so that you can talk to why it might not be a good decision and explain that to other stakeholders that might not fully understand the problem that they're working on. Yeah, you could tie it to longer term goals for the organization and come back. I'll be super intrigued to hear of someone or an organization where the engineers have said, you know what, we ethically we can't build this and how that played out. I haven't heard many people talking about it out loud yet. Yeah, and I think it's about choosing your audience. So if you have a C level person that you're trying to break the news to and say they're a CEO, for example, so they're worried about what the business looks like, how much money it's making, all you like, you just have to explain how if there was a new story about our company, building this thing, and this is the worst case scenario, and that's gone to press, and that might be enough to bring people along on the journey. So it's about understanding the people that you're talking to and how they kind of tick to make sure that they understand. Yeah, great point. And I think too, there's some executive people can have a gap in knowledge around, particularly ML and AI, and they put a barrier up first. But I think if you've built that relationship earlier, then you can give them the hard news at that point. So sorry. Hey, let's change tags a little bit. You founded Women in Machine Learning and Data Science, which in my head, I turn into Wilds and miss out the M completely. So that was about three years ago, you started, was it three years ago? And what made you go there? And yeah, so it was a fairly interesting story. So I had started at Eliza, I'd landed my dream gig as a data scientist. And I was in my kind of first few weeks there, we'd worked on a project, predicting, I think it was predicting like likelihood of a car spot in Melbourne CBD being available at a certain time of day. Obviously, the answer was nothing is ever available. This was, this was pre COVID. But yeah, our hypothesis around Melbourne's parking being hard was was definitely confirmed. But yeah, so we'd worked on a project, and my boss had asked me to write a blog about it, which I did. And I'm not, I'm not really a natural writer. And so there was a lot of pressure on me. And I wanted to prove myself I was kind of representing, not only representing myself, but representing the team that had worked on the project. So I was quite nervous about it, put a bit of effort into it. And it was really well received, it had been picked up on Twitter by the CEO of the city of Melbourne, and which had led to, it led to some talks with with their marketing team around potentially doing some work for them. So my boss was super happy. And, and then that week, I rocked up to a meetup that a colleague had been organizing. And at the end of it, it was essentially it was all men there, which is fine. And, and at the end, a guy had come up to me and Eliza team and asked about the blog that the guy in your team had written. And it was it was about my blog. And then in the profile picture at the time, I had short hair. So he the person's assumed that anyone with short hair is male. And anyway, so someone had someone corrected them and said, Oh, no, Fiona actually wrote that blog, and she's right here. And, and the guy just continued talking to everyone else apart from me, about my blog, about the models that I use. And it was it was a really jarring situation. Like I, I've come across lots of like, little misogynistic things that are kind of not not as shocking there, there may be more. And what's the word? Yeah, more subtle, but this, this is a real kind of slap in the face. So I got pretty upset. I went home in tears on the tram that night. And I was like, fuck this. This is a real problem. And it's Yeah, it's making me sad. But that next morning, I went into work and James, the CEO, who hadn't been at the meetup or had known that I was upset, but he sent me the link to the Women in Machine Learning and Data Science web page. And he said, Oh, this looks cool. And there's not a Melbourne chapter, you should think about starting one. And at that point, I kind of decided, yeah, I'm gonna, I'm gonna do this, I'm gonna turn this crappy situation into something really good. And yeah, prior to COVID, we were running monthly events. And with I think there's, there's well over 1000 members, but at each event, there would be maybe around 30 to 40 people would come along. And we'd have awesome talks. And there's a really nice community, everyone was supporting each other. And I don't know about you, but I've been to both mixed gender meetups, and women focused meetups. And then the, the the vibe in both of them is is very different. I feel like it at mixed gender ones, and the questions then are, I have always felt it's always just people trying to sound smart and kind of outdo each other. Whereas at women focused meetups, everyone's just there to learn and there's not this competitive edge. But yeah, COVID came along, and we have been kind of MIA for the past year. That being said, we we were hoping to have a in person catch up. And in the beginning of June, but we are in we are in Melbourne and the news is precarious. Yeah, it's a bit precarious. So anyway, long story short, watch this space, we are definitely having a welcome back party. And we're just trying to figure out when and where. Oh, send me an invite, or I'll find it and I'm definitely coming. When you said around the two different types of questions at the at the meetups, we had a similar experience. I run Women Who Code Melbourne, and we had Sam Aaron out from the UK who's written Sonic Pi, which is music based coding for kids to learn. And he ran a seminar for us. And it was so much fun. We're just making music with code. And at the end, I drove him to the airports, he was going back to the UK. And he goes, that was the best workshop I've run in forever. He goes, everyone was just there to learn and support each other, not to tell me why what I built was rubbish, and that they were smarter than me. Yeah, that's awesome. Sonic Pi is seriously fun to play with is the short take from that. Fiona, you're a role model in the industry. And I know of people who have moved into data science after hearing you speak or reading about you or hearing about you and just going, wow, this is a place for me, I can I can be in data science. So your passion for diversity is a bit contagious. Why do you think diversity is particularly important in data science and machine learning, I guess? Yeah, well, yeah, it's awesome to hear that firstly, before I jump into the question. When I became a data scientist, and I remember my first week as a data scientist, a colleague in the team said to me, okay, what is your career goals now? And because I've been so focused on becoming a data scientist, and that I wasn't sure what my career goals were, because I was just so focused on on becoming this, this, this person. And, and then, so from there, I kind of decided that what my goal career goals are, are to be someone visible within the industry, and to let people know that anyone can kind of do it, you don't have to look like a stereotypical tech person or a stereotypical corporate person, you can you can be authentic, and you can be yourself. And you can be a woman, and you can be queer, you can be you can be you can be you, and you can still find a place in this industry. And so yeah, really, really good to hear that, first of all, and why is diverse important? So, um, yeah, I mean, I mentioned the ethics things thing earlier, but, um, yeah, unless we have kind of diverse, diverse people, and diverse ways of thinking at the table, and in the decision, and in the planning process, and making decisions about how to build stuff, then we are going to build stuff that that can that go wrong. And so, so kind of making sure that machine learning and data science is accessible to everyone, not just technical people, or not just people with PhDs, to make sure that they understand the implications of the systems that we're building, so that we can we can build build things better. It's really important. And not only that, but more diverse teams, they have been proven to create more creativity and innovation, and they become better problem solvers as a team, they build better products. And if you're talking to a CEO, you can tell them that they make more revenue. And, yeah, there you go. And, and also people people want to work and well, a lot of people will want to work in more diverse teams. So, yeah, it's kind of a no brainer for me. And, and I think I think that it's something that that everyone should think about and, and, and kind of do a bit of research and look at the stats, because the stats are there. That's, um, as you were listing off all the reasons why it's amazing. I was thinking, why would anyone not want to do this and have a really diverse team? I guess, is it because it's hard to find people? Why do you think people? Why do you think we don't have more diverse teams at the moment? And yeah, I mean, there's, there's lots of layers to it. But maybe the people who might be a little bit resistant to diverse teams, they might be the ones who are benefiting from not having diverse teams. And so that would be my initial thought. But I suppose taking it taking it right back to the beginning where we're telling we're telling little girls at school to like play with dolls, we're telling little boys to play with Lego, to a certain extent. I mean, that was certainly the case for me growing up, I think we're coming a long way, but there's still a lot of gender stereotypes in force from a very, very young age. And that kind of gets ingrained. And there's this general kind of narrative that's funny and cool to say, and that's that mass is really hard and complex. And I just don't understand that. But and it's not it's not hard. And I think I don't think we've done a great job at teaching it. I don't think we've done a great job branding for it. And we're quite, we're quite blessed in that it's it's kind of a little bit cooler to be a geek now. And but I think we do have a lot of work to go to, to make mass accessible and kind of break away from this idea that it's hard, and it's nerdy, and I can't do it. And so yeah, it starts way back then. And then real kind of the patriarchs, patriarchal system will come in and kind of lead to other other reasons why why we don't have gender diverse teams. And that's, yeah, a number of different reasons. There's, there's more men, there's currently more men at the top. So they might be hiring more men, and workplaces, they might not be offering flexible arrangements to maybe make. Yeah, to encourage more women in the team. And yeah, there's, there's lots of reasons pointing to why it's a problem. And but yeah, we should definitely start focusing on what we can do about it really. And how to solve it. I heard a talk quite a few years ago, and I'm pretty sure it was Lisa Harvey Smith talking about we do, or enforce a gender bias around maths from age two, apparently. So and the example was, if you're going for a walk with a two year old boy, and you come past a garden, you'll go, let's count the flowers. And you sit there and you do some maths. And when they were watching, it was a huge study. And I'm paraphrasing this down to nothing. And when you went past with a girl in a pram, the questions were, look at the pretty colors. So already, there was a, you can do maths and numbers, because you're a boy, and you're a girl do the pretty things. And it was from age two. And that is just, yeah, it's terrifying. So terrifying. Sorry, you go. I was just gonna say, and even then, when we when we are teaching kids about maths, and there's always that kind of age old saying of, why do I need to like, differentiate, I'm never going to need to differentiate a curve in my life. And then, and I thought that too, I'm like, but I, I enjoyed doing it. And I was somewhat, somewhat good at it. So it kind of kept me going. But equally, I was like, Oh, I'll never use this again in my life. And then I started learning about data science. And I was like, oh, shit, this is where all the stuff I've learned in maths is coming into into play. So and I think, making maths more applied. And yeah, from from a young age, I think would be really helpful to is an interesting point. Because if you think, cooking is chemistry, right? Every part of cooking is some form of science. And yet, you don't get people saying, can't cook because I didn't do chemistry. We kind of with maths, we go, we do it, we can't do data science, we can't do anything, because we didn't do pure maths one day. And it doesn't quite hold true. So I think your point about branding is amazing, and it should be looked at. But also, how we teach it is so removed from context. I wish, actually, this is a total teaser for Friday's expert I've got coming in, is Dr. Linda MacGyver. And she's actually built up a program around teaching data science in a practical manner. So and it's had great feedback from students and educators, because the children we have are not stupid, and not to be hoaxed by some gimmicky game. And if you teach them data science and practice, they are loving it. I'm way off on tangent there. Have you seen any organizations do something kind of creative around getting more diverse people into their teams? I know, and that's a bit leading. I know Mental Group have done some good stuff in this space. Yeah, so I can definitely talk to Eliza and their policy, and they would have lots more policies since I worked there. But they would only hire women or gender diverse people into associate data scientist roles. So they recognize that there's a massive kind of talent pool within people who want to come from academia, people potentially with PhDs that are super smart, but just lack industry experience. So yeah, James made a very clear and conscious decision to save those associate data science roles for gender diverse people or women. So yeah, I think that was awesome, just hearing him, because quite often things like quotas or like only opening certain roles to certain people, it will come with some pushback and some resistance. But to see him just proudly say one day, yeah, this is what I'm going to do and do it. It was a really cool moment for me to see, and it's definitely something that I've carried on when I'm hiring associate data scientists. And any role, yeah, it's definitely front of mind when I'm trying to source talent for my team. What a powerful move, and what a way to set culture in an organization. Yeah, definitely. Unashamedly. Hey, I've got, well, actually, I've got two final questions. I'll do the shorter one first. There's a lunar eclipse tonight. Are you going to be watching it? Yes, I think I will. Hopefully, the clouds clear up so we can see it. That would be pretty cool. I'm super excited for that one. But my actual final question was, what do you think the biggest myth in data science is? Yeah, so similar to what we just spoke about maths, but there's this idea that it's really hard, and you need a PhD to understand it, and that you need to know everything before you can do something. I think that's definitely a myth that I've seen. And I think it really comes down to the way that we are communicating the work that we're doing. So it's something that I'm super passionate about, is being able to explain what I did and what the system that we built did in everyday terms, so that every single stakeholder that is involved in the project understands what's going on and what all the different parts means. But yeah, it's really about getting the communication right with the stuff that we're building, and I feel like we're not quite there yet. But getting closer every single time. And I like it. The communication piece is powerful. I'd like to thank you for joining us today, Fiona, bright and early. Pleasure, yeah, it's very early. It so is. I always love hearing you talk and your insights into the data science world. On Friday, as I mentioned, we've got Dr. Linda MacGyver coming. She's unbelievably amazing and intriguing and interesting. And she's going to talk to us about how we can start raising heretics, which is what we should all be doing, I think. I'll see you then.