Paul McLachlan: [00:00:00] I struggle a lot with imposter syndrome, and I think the thing that I want people to remember is you don't know the story of the person who is sitting across from you or sitting at the desk next to you. We assume everyone is gone from success and peak and peak and peak. And there's never been a value. There's never been a setback. That's just not true. Everyone has valleys. Everyone has setbacks. Everyone you're interacting with from the CEO to to your professor is a human being just like you. Just keeping that in mind to me has really helped. It's not solved, but it's helped some of the imposter syndrome. We're all humans. And I find a lot of inspiration and a lot of creativity in burning people's story. Harpreet Sahota: [00:00:54] What's up, everyone? Welcome to another episode of The Artist of Data Science. Be sure to follow the show on Instagram @TheArtistsOfDataScience and on Twitter @ArtistsOfData. I'll be sharing awesome tips and wisdom on Data science, as well as clips from the show. Join the Free Open Mastermind Slack channel by going to bitly.com/artistsofdatascience.Where I'll keep you updated on biweekly open office hours. I'll be hosting for the community. I'm your host Harpreet Sahota. Let's ride this beat out into another awesome episode. And don't forget to subscribe, rate, and review the show. Harpreet Sahota: [00:01:40] Our guest today has over a decade of experience applying his knowledge and expertise to the great trifecta of industries. I'm talking corporate and academic, all mixed in with some entrepreneurial endeavours. Harpreet Sahota: [00:01:52] He's a passionate communicator and Data science educator who strives to make our art form more accessible and understandable so that everyone can make better data driven decisions. He successfully coached and managed cross-functional teams of all sizes to help them translate their ideas into successful businesses and improve their go to market strategies. He studied at some of the top universities in North America, from Columbia to George Washington to the University of California at San Diego. He's gone on to earn a PhD in computational social sciences from UC San Diego, where the bulk of his research focused on applied statistics, machine learning and causal inference. He's also published in multiple academic disciplines and has a plethora of patents and if all of that wasn't impressive enough already, he's contributed to research efforts at the San Diego Supercomputer Center, Emory University and the University of Michigan. His contributions and expertise have led to numerous startups and nonprofits inviting him to serve as a mentor or advisor, both in the San Francisco Bay Area and abroad. He's currently the eye research leader for the Consumer and Industry Lab at Ericsson Research and was previously principal data scientist at Ericsson's Global Artificial Intelligence Accelerator, where he was responsible for formulating Data strategy and influencing Data monetization for a global telecommunications company that powers 40 percent of the world's mobile Data. So please help me in welcoming our guest today, a man who has provided technical and innovative leadership around the world. Dr. Paul McLachlan. Harpreet Sahota: [00:03:22] Paul, thank you so much for taking time out of your schedule to be here today. Man, I really, really, really appreciate it. Paul McLachlan: [00:03:29] I'm really I mean, I'm I'm I'm happy to be here. And it's a good thing that this is not a video podcast, because after hearing that intro, I am blushing beet red. Harpreet Sahota: [00:03:38] Oh man - Well, you know, you've accomplished so much in your career. It was really, really hard for me to write that intro because I had to choose between what to leave in and what to leave out. Talk to me about your path into Data science. What sparked your interest in the field? Where did you start and how did you get to where you are today? Paul McLachlan: [00:03:57] I mean, I think that's a really good question and one that doesn't really have an easy answer. So I don't think I didn't want to be a data scientist when I grew up as a kid. Paul McLachlan: [00:04:06] And in fact, if anything, I'm a high school dropout, so I have my GED. And I think I'm one of the very few people who have both a GED as well as a PhD. And I would say, you know, what really sparked with me is I did my undergrad at Columbia University in New York and they require every undergraduate to take a math class. And I put it off until my very, very last semester. I was there because I was I was genuinely afraid of making mistakes, of looking stupid and doing badly, of getting a poor grade. And I am incredibly grateful to my graduate teaching assistant who was leading that class, because I went on the first day and just said, I am scared. I am nervous. Can you help me? Can you help me remember math? Because I as I mentioned, I was a high school dropout, so I went through and I was in office hours, three hours every week. And really I was that one-on-one experience, that one I thought really connect me with the power of of math. At that point, we just called it statistics. It wasn't data science, but really got me understanding the material. And I think from that, the thing that really got me excited around Data science or statistics and I hadn't really understood what math is, the ability to answer questions. So I'm a really curious person. And I thought, oh my God, there's this technology or this technique that you can use to test things, to understand questions. That is just a coolest thing I've ever heard of. And that fuelled me through graduate school and now in my career, because I just think of this as a technology to answer questions in a rigorous way. And I, I, I just think that's the coolest thing and that's what motivates me. Harpreet Sahota: [00:05:50] That's awesome. And I think a lot of people actually breaking into the field, they tend to be afraid about that math and all that technical stuff as well. So it's really interesting to hear your perspective on how you also were very fearful of digging deep into that, but eventually went on to achieve, you know, PhD and in a very quantitatively rigorous field, if you wouldn't mind, can you share kind of what got you out of that mindset of, you know, being fearful of math, not wanting to, you know, look, quote unquote, stupid. Talk to us a little bit about that. Paul McLachlan: [00:06:19] Oh, I mean, I'm I still think I'm pretty worried about looking stupid. I think. I think that's pretty natural for every person. You know, we always second guess the questions we ask. I think a lot of it just came from one solely realizing that for many topics, certainly not all of them, but for the ones I've studied, you know, I I actually am the expert in the room, and that might not always be true. But I think a lot of that comes from learning to trust yourself. Getting validation, but also realizing that everyone comes from a different angle. And even if the question sounds silly to someone else, you had that question. You had that thought. And so that means that the person communicating, the person doing the the teaching or the person leading the project didn't communicate or impart enough information for you to know the answer to that question. One of the things I have tried to reframe my thinking around is not who's smarter or who's less smart, it's who's communicating and that I get all the information I needed. So I tend to think of now when I ask questions as trying to make sure I'm getting the information I need. And I feel like that maybe makes it less evaluative and makes it really just collaborative. And I think that that takes a lot of the, at least for me, the mental stakes out of potentially asking stupid questions. But I think the adage that there are no stupid questions is is really true. Harpreet Sahota: [00:07:42] Kind of refreshing for me to hear you say that because, you know, you've gone on to accomplish some amazing things in your career. And I know that myself and many other data scientists also will sometimes feel that kind of bit of that imposter syndrome. And to hear somebody that's at the top of the game say that they still kind of struggle with that and feel that it just makes it makes you not feel as bad about it, I guess. Really. Thank you so much for sharing that. Where do you see the field of Data science machine learning and A.I. headed in the next two to five years? Paul McLachlan: [00:08:16] So I work at Ericsson. And Ericsson is the world's leading telecommunications company. Paul McLachlan: [00:08:20] So as you mentioned in our introduction, we power 40 percent of all of the world's mobile Data is traveling across our networks in a given moment. And what we're really excited about and what I'm excited about is 5G. And I think that the connection between 5G and data science is not immediately obvious. But for me, what I think I want about is 5G reduces latency. Paul McLachlan: [00:08:42] So what that means practically is that the lag time between a Data, you know, some data being recorded and being able to process it in the edge or at the edge or in the cloud is significantly reduced. So a lot of what we work with when we're building machine learning models is historical data. And what I am really excited about over the next two to five years is benefiting from that significantly reduced latency to work with real time data. So I think that that will improve Data precision, it will improve Data accuracy. And also is really interesting implications for privacy. But for me, I think the real shift in the near-term is going to be towards thinking of real time data and the type of systems we can build to work with real time data rather than trying to build systems to work more and more with larger a larger historical data sets. Harpreet Sahota: [00:09:38] So you mentioned privacy issues and privacy concerns in there. What do you think will be the biggest area of concern for the application of A.I. in the next, say, two to five years? Paul McLachlan: [00:09:49] I mean, I think that the conversation around algorithmic bias or algo - and trust in algorithms is going to be increasingly important. Paul McLachlan: [00:09:58] I think that we have seen some societal changes where we are giving more and more emphasis to machine learning, but we haven't necessarily brought people on board in that journey. So people feel or worry that a lot of decisions are being made by algorithm. I think privacy and trustworthiness and A.I. is a critical issue and something that I personally take extremely seriously. And I think that Ericsson and really most companies as well take pretty seriously where I think that we can really work to improve, where I think we are working to improve at Ericsson and hopefully as a discipline is making machine learning much more accessible and a sense of using explain-ability or explain-ability techniques. Because right now, I think a lot of people are concerned that algorithms are black boxes and that means that they don't know what is happening outside of the algorithm and maybe the data scientists themselves don't know. And I think being able to offer explanations and interpret ability will make those algorithms much more trustworthy and by trustworthy, I mean at the societal level. And I think that that is going to be an incredibly important focus for A.I. in data science. And as regards kind of privacy and trustworthiness going forward in this vision of the future you have there. Harpreet Sahota: [00:11:22] What do you think will separate the great Data scientists from the good ones? Paul McLachlan: [00:11:28] I think willingness to actually ask good and difficult questions. So that to me, I think touches on the imposter syndrome you asked. I think one of the challenges for Data science is. We both have to have an incredible amount of domain expertise as Data scientists, so we need to be on top of the literature because it's constantly evolving and there's always new things to learn. We also are now starting to embed ourselves into industries. So that requires us to have subject matter expertise, both as Data scientists and for the industries in which we are working. So I think someone who can combine those two together creatively and then speak passionately about that combination is really well, you'll see kind of great Data scientists distinguish themselves from good ones because we need to make sure that the tools we are using bring the value to the industries in which we are working. It also means that we have a lot of work to bring our non-technical stakeholders along the journey with us because we can build the best and most innovative cutting edge algorithm. But if our sales team doesn't feel empowered to talk about it, that can be a challenge. Or if our stakeholders don't understand the research and the innovation that went into it. That can also be a problem. So I really think a great Data scientists brings domain expertise and machine learning, subject matter expertise in their industry and an ability to bridge the two. Harpreet Sahota: [00:12:54] That was really well put. Thank you so much for that. Harpreet Sahota: [00:12:57] So I've seen Ericsson mentioned in the news and the new post I linked in recently with respect to the efforts of the 400 volunteers who took part in the White House Office of Science and Technology COVID-19 open research dataset challenged for information retrieval problems that we're using NLP. For those people who are tuning in them maybe aren't Data scientists or aren't familiar with the concepts. I can quickly give a high level overview of what information retrieval is, Paul McLachlan: [00:13:24] Of course, so information retrieval is Paul McLachlan: [00:13:26] pretty much like building a search engine. So if you have a group of documents in this case we were looking at when we started in March, twenty thousand academic papers have been published on COVID-19. Now, in May, I believe there are seventy thousand papers in that same corpus. So the goal is how do you retrieve information or how would you search through those documents? So the goal for that type of data science is to match understanding of a question. So I want to know more about this and then mapping also, which documents talk about that that specific topic. Harpreet Sahota: [00:14:02] And how about NLP? Paul McLachlan: [00:14:04] I mean, in this case, you know, NLP - natural language processing - is trying to understand meaning from words and and documents and things like grammar, sentence topics. And so when we are working with trying to retrieve information from written papers, we use and all techniques to understand what those papers and what that document is, is talking about. So that gives me like topic modeling where you're trying to identify, you know, which is this paper talking about to, you know, genome sequencing or is it talking about ethical considerations of medical care? And that's where NLP comes in. We're working with the written word. Harpreet Sahota: [00:14:40] Awesome. Thank you so much. So that probably covers the two non-data scientists who aren't - I'm sorry, the two non Data scientists listening to the podcast. Which is probably my wife and my grandma. Now that we've got now that we've got that covered for them. Could you talk to us a bit about the innovative solutions that Ericsson data scientists developed for this challenge? Paul McLachlan: [00:15:00] Yeah, I would be absolutely thrilled to because I should say I really was an incredible honor getting to work with this. And I think I want to set a little bit more of a picture around what we did and how we came together. So the White House Office of Technology and Policy put out a call for tech companies to get involved in this research challenge. So, as I mentioned, there is an incredible and growing corpus of medical papers that discuss COVID-19. And, you know, even when we started 20000 thousand papers, it's 200,000 pages or so, give or take. That's more than any human could read, probably in their lifetime. But the really the ask was we need to get information to medical professionals now. Not in six months. Not in six weeks. Now, they gave a three week, sorry, four week turnaround time to go from this challenging post into submitting a response. And what I found was really incredible and really to me speaks about our our culture at Ericsson. Within 12 hours of this challenge being posted, we had gotten agreement from our global CTO, our head of North America, or head of Ericsson research in Silicon Valley, to go ahead and engage in this challenge. And we were expecting when we sent out the email to call for volunteers to all of our employees in North America. Paul McLachlan: [00:16:16] We were expecting or hoping for 20 to 30 volunteers and we got nearly 400, which was incredible, but also terrifying because, you know, I had to build a organization structure. We had to come up with a communications structure, a team structure in place. So what I felt was really incredible is we worked really quickly, which can offer. The challenge for Data science projects, because there's so many different techniques you could apply. There's so much learning to be done. And we also really focused on making the results accessible to non Data scientists because the ultimate goal was to make information accessible for the medical and the policy community. So we put a lot of focus on visualization as well as you and you axe. And those are two elements. I think that we don't always emphasize as much and Data science teams because they can sometimes be a disconnect between the technical work and the end user. But for us, we had the end user really clearly in mind and what their needs would be. So that was really for me. One of the coolest aspects is getting to make sure that the results we put together, the models we built were useful for the medical community. Harpreet Sahota: [00:17:36] Are you an aspiring Data scientist struggling to break into the field within Check-Out dsdj.co/artists to reserve your spot for a free informational webinar on how you can break into the field that's gonna be filled with amazing tips that are specifically designed to help you land your first job. Check it out. Harpreet Sahota: [00:17:57] dsdj.co/artists, Harpreet Sahota: [00:18:01] 400 people volunteer, but they all Data scientists or they kind of a cross-functional team, Paul McLachlan: [00:18:06] Completely cross-functional. Paul McLachlan: [00:18:07] So we had everything from some of our most senior and experienced Data scientists. Paul McLachlan: [00:18:11] So I should also note, even though the challenge or the opportunity went out to employees in North America, we had volunteers in Canada, the United States, Sweden, Ireland, Brazil and India. So it became a completely global response. But we had everything from because, again, we had to make sure everything was valuable for medical practitioners who may be in the field, who may not have 10 hours to sift through all of the papers, even the ones we identified for them. So we had everything from experts working on summarization techniques and applying those to technical and non technical writers who could help us document our code, help people understand how to use it. We had visualizer as we had Data engineers, we had project managers. It really was an incredible coming together of a variety of different skills. And I think that's also to me highlights that Data science is really a collective endeavour. And we even Data scientists, teams, even the most skilled and successful data scientist, is going to have to be able to successfully work with technical stakeholders, non-technical stakeholders. And that's why, you know, I think it's it's incredibly critical to always be able to communicate across those different types of understandings of the subject matter. Harpreet Sahota: [00:19:31] It's really, really interesting. Are you able to share at all kind of the with the end product looked like once it was in the user's hands. Paul McLachlan: [00:19:40] Yeah. I mean, they're available online. So I would be happy to share the link with anyone. We can post it in the podcast notes. But I'd say one of the findings that came out that I thought was really fascinating. So we applied some signal processing techniques to genome data and we actually wrote a white paper about the technique we developed, which I would love to share with the audience as well. But even using that technique, we are able to identify, for example, about the virus samples we had access to the original sample started in China. So we had the time, date and location where the genome was collected for the virus and then spread into Western Europe and the United States relatively simultaneously. And the samples we had from Brazil looks like the actually might have come from the United States. And it always looks like the virus spread from China to Western Europe and United States simultaneously. And then from the United States into Brazil and South America. So we were able to using signal processing techniques, which are very part and parcel of computer science, to actually understand the spread and evolution of coving 19. Harpreet Sahota: [00:20:52] That's a really, really fascinating. Yeah, definitely, I'll get those links from me and I'll be sure to include that in the show notes so that our listeners can go check that out. So that was a really interesting finding there about the spread of the virus. What are some of the other interesting findings that you guys kind of got from this project? Paul McLachlan: [00:21:10] I mean, so one other aspect we're looking on now is, what did we learn about how to build natural language processing models and information retrieval? And so we have a variety of different avenues of research going on there, which I would love to follow up with, when we have papers drafted and we can share. Harpreet Sahota: [00:21:28] That is so cool. Awesome. Hey, so what are some of the ways you see Data science and A.I. helping fight the COVID crisis going forward? Paul McLachlan: [00:21:37] I think that that is an incredibly important question because I think everyone's life has been impacted by COVID-19 some tragically, whether illness or loss. And and some just in terms of changing of life or being in a shelter in place. And I think what we have to do is to minimize the human cost, that societal cost using data science and A.I. Paul McLachlan: [00:22:02] And I think that's first and foremost on everyone's mind. We also have to do so ethically as well. So I think there's a couple of areas I can see, and these are just observations I've had from reading the press. And I don't have any great expertise here. But let us hope very soon that a virus, a vaccine is developed. And once that happens, we have to make sure that we have supply chains to manufacture that vaccine at scale. And that is a whole domain for a AIs, Data science in terms of supply chain optimization and manufacturing optimization. And I think that that will be a critical area for people to think about. The other is that we then have to make sure that we can get that vaccine to all people. And that has, again, a very important supply chain and logistics problem. And then also, we have to then work with communities and understand communities around communication and effectiveness, because we have to discuss why this vaccine is important, why they might want to receive it, or why they need to receive it. Paul McLachlan: [00:23:11] And that is, again, a very critical space for Data scientists to learn how to communicate effectively, both in terms of studying and messaging, as well as being communicators themselves to the communities in which they live and they are working. And I think you could go on with a bunch of different examples in this space. But I think that you can also add in even the research that's ongoing to find vaccine candidates or therapeutic candidates. And from there, I think the number of ways data scientists can get involved is pretty much limitless. I would say, though, the one piece that we always need to be mindful of as data scientists is to make sure that our work is embedded and anchored to a real stakeholder need. And the sense that I think it's very valuable for us, particularly as data scientists, to be humble about our subject matter expertise. And very few of us are trained as epidemiologists or as medical researchers. And so it's very critical that we stay close to people who do have that subject matter expertise and even that modeling expertise because we don't want to get too far out in front of our skis. Harpreet Sahota: [00:24:23] Thank you for that. That's also very, very important message and also given me and our audience a lot to think about. So thank you so much for that. Harpreet Sahota: [00:24:30] Congratulations on your new role. AI Research Leader for the consumer and industry lab. So can you tell me a little bit about how the consumer and industry lab fits into Ericsson? Paul McLachlan: [00:24:42] Yeah. So Ericsson, as I mentioned, is the world's leading telecommunications company. And I would say many people when I mentioned that I work at Ericsson are not familiar with what we do or what our what our company is. And often I'll get asked, do you guys still make cell phones or something along that line? But Ericsson is really the world's leading telecommunications company. And so that means that we build, design, service and install many of the world's mobile networks. So here in North America, if you're using your cell phone to listen to this podcast, it's very likely that you are connected to an Ericsson managed network and using Ericsson or radio equipment to connect. So we are very core and essential to the world's connectivity. So because of the importance of research to Ericsson to make sure we are always at the forefront, we have thousand R&D staff around the world. So R&D is incredibly core and essential to what we do as a business. And the way consumer industry line fits into that is we are the voice of consumer and industry inside Ericcson Research. So we help understand and work with stakeholders to understand what consumers, industry industries will be doing with mobile networks. Five years in the future. 10 years in the future. And then we help guide our R&D teams to make sure that we have the technology, the know how to make that technology vision possible. So, for example, we recently released a report around the Internet of sensor's, which is the idea that by 2030 we forecast not only will people have X or VR glasses, but they'll be able to interact digitally with all senses. So touch, smell, taste through haptics, through a human brain interfaces and the like. And so it's really incredibly, very visionary insights. And then Eriksson's R&D team will be guided in part by that. That vision is setting the consumer industry lab does in terms of the orientation and direction of the research we engage in. Harpreet Sahota: [00:26:56] Wow that sounds really, really fascinating. So, you know, you've worked on several XR/VR related patents. Can you first before anyone who's not familiar with those terms is kind of defined what, XR and VR are and maybe share with us What aspects of XR and VR are most interesting to you. Paul McLachlan: [00:27:15] Yeah. So VR. Just is short for virtual reality and XR short for extended reality. So in virtual reality, I think the idea is you would wear a glasses or a headset and the environment you see through the glasses would have very little to do with the environment in which you are in. So, for example, you might be in your home, but you would use virtual reality to be at a concert or to watch a movie or to travel to a place you'd never been to before. Whereas extended reality, the idea is to augment, to improve or to give you additional information in the environment to what you already are. So, for example, it might help give you directions as you're walking around the city you've not been to before and might help guide you to a bus stop. Or maybe you would watch a movie or a video while you're sitting on the couch through XR. So the idea is, is with XR is that time essentially just augments the road you're in already. Where as virtual reality kind of replaces it. And I think what I am really excited about, maybe Excite isn't the right word, but one thing I think is really important from a research angle is security in XR and VR. And that can mean everything from making sure that the types of content people see is safe for them. How do we also make sure that people are not manipulating the content people see? So if you think a deep fakes, that could be a concern that comes up at XR and VR. And we are working on solutions around that as well. I think that the layering in a physical space, which is what happens with extended reality, poses additional implications for security that are very important from a research perspective and very rich resource. From a research perspective. And give us a lot of opportunities to innovate, because as we as this technology matures, it's critical that we put ethics, safety and trust first and foremost in our research and development. Harpreet Sahota: [00:29:19] You mentioned something that I kind of flipped the light in my head about concerts, movie experiences, things like that going forward in this social distancing world. I don't know how long this is going to be a thing or how long this is going to impact our whole world for. But a lot of applications for that with you know, concerts now being fully VR, the movie going experience being fully VR. Like what do you see virtual reality kind of having an impact in the mass consumer market. Paul McLachlan: [00:29:50] I mean, I think that's that's a real and important use case for this technology. I think that the fundamental piece we have to remember with kind of concert going or watching movies, these are fundamentally social and shared experiences. So it's nice. You know, you can watch a movie now with Netflix. But why we go to the movie theaters where go to a concert is to share that experience. Many cases with friends, you know, even myself, I don't know why, but I don't know if I've actually ever gone to a movie on my own. To me, it's something that I always want to share with someone else so we can. Now, you know, there's so many Zoom parties where you might watch a video or a TV show together with your friends, but it doesn't have that same immersive ness that go into the concert. Going to the movie theater with your friends has. And I think that XR can help bring that immersiveness, but also that shared element together. So it's not just, you know. Paul McLachlan: [00:30:51] I have a little window on my laptop and we're or chatting and as as a movie is playing through Zoom or through any other messaging app. We are at that concert together and that is, I think, captures and a much more fundamentally human way. I know what these experiences mean for us as as social animals. Harpreet Sahota: [00:31:13] It's really, really interesting. By any chance, have you seen the show on Amazon Prime called The Feed? Paul McLachlan: [00:31:19] I have not seen that, no. Harpreet Sahota: [00:31:21] Oh, man. It's pretty much the future that you're describing. That shows a fictionalized version of it. And I think you'd find it really interesting if you get a chance to check it out. And I'll be sure to post a link to that show as well for listeners here. But that's really interesting. I think you'd enjoy it, especially with your research and everything you've talked about here. Thank you so much for that information there. Harpreet Sahota: [00:31:45] I wonder if we can switch gears a little bit and maybe pick your brain on, you know, some some career tips for Data scientist. You know, as someone who's the first data scientist in an organization and who wants to spread a message of data driven methodology. What are some challenges you foresee some one like that facing? And can you shares some tips on how to overcome them. Paul McLachlan: [00:32:11] Yeah, I can. Based on on some personal experience, as well as conversations with friends and colleagues, I think particularly for, you know, really smart and bright people graduating with a bachelors, masters, PhD in and Data science or statistics or computer science. Paul McLachlan: [00:32:31] We've gotten or you get really used to being around people who understand the concepts and vocabulary you're using and also why they're important. And what then happens, though, is when you're hired into a company, you maybe for the first time or for the first time in a long time, an environment where people don't share that same understanding as you do and also don't understand the value that some of the work that you are doing brings to their organization or to their own team. And so I think what can often happen in those senses is we don't have as much practice in explaining. Paul McLachlan: [00:33:10] I don't want to say in layman's terms, because I think that that is actually underselling the importance of this type of conversation. But in non-technical terms, why the work that you are doing matters and what value it brings to stakeholders, to the organization, to a bottom line? And I think that that is actually super critical. And I think that it is something that is very, very difficult to do. And it requires a lot of practice. And it also, I think, requires data scientists to be proactive and reach out to non-technical parts of their organization for coffee, for lunch, for, you know, for a virtual hangout nowadays or social distancing to understand what it is that the teams you're interacting with are trying to accomplish and how Data science can help improve the work that they are trying to do. But it also requires a lot of humility because it requires starting from a premise that you have some incredibly valuable and important technical skills, maybe ones that no one else in your organization has, but particularly for a Data scientist who's just starting out. It also requires a lot of humility and acknowledging that don't have the subject matter expertise yet that the stakeholder you're interacting with does. And so while you might know the very best and very most cutting edge solutions to a problem, you might not understand how to apply those tools to that problem at hand because you don't maybe have the subject matter expertise to understand all of the dimensions and pros and cons and all of the attempts that have come for you to try to address or or or improve something in your organization. So it really starts from a position of humility. And I think that that can go much further for Data scientists than always trying to be the smartest technical person in a conversation. And that's a really hard lesson to learn and one I've had to get reminded to me many times. Harpreet Sahota: [00:35:13] Awesome. Thank you. Yes, I was going to ask you for my next question is what are what do you look for in a data scientists beside those those technical skills? I mean, you mentioned humility as being one of them, but what can we do to kind of cultivate that quality within ourselves? Paul McLachlan: [00:35:29] I mean, for me, I would say part and parcel of humility also comes curiosity. And why I say that is I think curiosity means to me when I interact with humans are curious and means that they want to know more. And to me, that's just just an acknowledgment of the limits that any individual human can know. We're always going to have part of the picture. We're not going to have the. For answer or maybe some of us are very, very lucky to be. And some of us are fast smarter than I. But they will, maybe some will have that full picture of knowledge. But I really - when I am hiring people, when I'm talking to people, I really resonate with people who are curious. And that means people who are willing to go out and talk - when they're on the job, talk to people who they might not have interacted with before. Ask questions, even if they might might maybe sound very basic, because to me it indicates they really want to understand. And through that understanding, it's going to shape and guide how they apply tools and Data Science is a set of tools that we have to solve problems on. Paul McLachlan: [00:36:36] So I really react very positively when someone kind of expresses curiosity in a Problem-Solving mindset. And I don't know how to inculcate that. But maybe one good approach or just as a practice is if you're inside of a company, try to reach out and have virtual coffee or conversation with someone whose work you don't fully understand or what they're accomplishing. You don't know. You haven't had a chance to interact with yet and just chat with them for fifteen minutes, 30 minutes. And if you're looking for a job, go to open houses, go to virtual events or learn, because that's also really going to help guide your career, because it's I would say it's very valuable to understand a particular industry or some of the questions that come up in industry. And when you're graduating from a Data science program, you still haven't tried in all the industries that are out there yet. So going to open houses, going to startup events, going to meet UPS is going to help one, fuel your curiosity and teach you a lot about what different industries are working on and what they're trying to solve. Harpreet Sahota: [00:37:47] I really like that advice, as well as how to get exposed to different industries and how to understand different industries. You mentioned networking, speaking with people. What else can we do to kind of gain some industry knowledge? Some industry awareness is. Do you have any other tips related to that? Paul McLachlan: [00:38:09] Read a lot. And I can always say these are things that I try to do. I don't know if I'm successful at these yet, but I try to read a lot in the sense of I read a lot of nonfiction. I also read a lot of fiction because I think to me it gives me a window into what different people or different problems that might crop up are. And I just find that it inspires my own creativity. But this is what works for me. I think what might work for it, for everyone else's is really going to depend upon you, your personality and what your interests are. But I think if you keep from a mindset of of humility and curiosity, it's a pretty good foundation to build on. Harpreet Sahota: [00:38:48] That's great advice. Harpreet Sahota: [00:38:49] I always recommend to my mentees that if they are wanting to know about a particular industry, the best thing is, like you said, read, but just read case studies, because by reading case studies, you'll be able to get exposure to the vocabulary and the terminology that is in that space that can then direct your research efforts in a more concentrated way. Paul McLachlan: [00:39:11] That's great advice. And it also is just so helpful even to learn the vocabulary and what's, you know, what solutions people have tried before because very few industries are brand new. And so they've gone through different different sides, different technologies, different approaches. And so particularly when you're communicating with people who've been in industry for a long time, they might view Data science, rightly or wrongly, as just another, you know, flash in the pan or a new trend. And so if you can articulate to them that you understand the industry, some of the history, you won't go so much further, I think. And being able to persuade people about your approach. Harpreet Sahota: [00:39:52] Do you have any tips for Data scientists to find themselves in a room full of executives and they need to communicate their findings or communicate their ideas? Paul McLachlan: [00:40:03] I can only share what I try to do myself. I don't know if these are successful because I think that any good data scientist know this is hard to know the counterfactual here, but what I what I find when I'm listening to a presentation, really powerful to me is when someone connects it to what that stakeholder is trying to accomplish. And what I mean by that is if you're interacting with your head of sales, what they're probably less interested or they might be at a personal level and some of the technical work. But they're you know what they're judged by, what their assessments are. How much do they sell? And so I think communication is very powerful, very effective when you tie it to what that person or what the audience needs to know. Or why they would be excited about the work that you have done so for Data scientists, a proper way means if you're interacting with your head of sales, maybe they're a little bit less interested in the architecture for your deep neural net. And there might be more interested in how the results in the model can make them more effective or make their teams more effective. And so that can be a very delicate balance, though, because you don't want to completely blackbox your model or to undersell the the effort that you put into building that model. But I think that tying that to what does an audience need to know? And why are they. Why would they be excited about the work that you've done is a really good frame of mind to keep when you're putting together a presentation or communicating. Harpreet Sahota: [00:41:40] What's up, artists? Harpreet Sahota: [00:41:42] Be sure to join the free, open, Mastermind slack community by going to bitly.com/artistsofdatascience. It's a great environment for us to talk all things Data science, to learn together, to grow together. And I'll also keep you updated on the open biweekly office I'll be hosting for our community. Check out the show on Instagram @TheArtistsOfDataScience. Follow us on Twitter at @ArtistsOfData. Look forward to seeing you all there. Harpreet Sahota: [00:42:10] Thank you for that. So what's your take on what it means to be a thought leader in Data science? And how can one go from aspiring Data scientist to becoming a thought leader in Data science? Paul McLachlan: [00:42:26] I mean, I realize there's a certain irony in me sitting down with you and doing a podcast and being a little bit also dismissive of the idea of being, a though leader in Data science. So I want to acknowledge, you know, some of the irony there. I think the real challenge is that my thoughts around being a thought leader are really focused on what is the value of the communication you bring. Paul McLachlan: [00:42:52] And I think anyone who is on LinkedIn has seen a thousand or one posts about Data science. But what I think really cuts through what is this signal to the noise is where you are communicating and sharing information that is valuable. And then it's as pitched at a level that makes the work you've done accessible. And I think people take a lot of different approaches to this, because what I mean by accessible can be very different depending upon your audience. So if you are a NLP expert and you want to be a thought leader on an LP, accessible might mean talking to other people who have very deep domain expertise in NLP. So maybe you don't need to be as mindful around some of the jargon, some of the framing, some of the motivation, because you can assume very clearly what people know. But if you're trying to be a thought leader in data science for executives, I think that communication will be very different. So I think the advice I might give is to be very mindful of who you are speaking to, what they know, what they don't know, and how the communication you are bringing is valuable to them, to whatever whatever the audience might be. Harpreet Sahota: [00:44:00] A big part of our audience will also be people who are breaking into the field. Transitioning into Data science. Harpreet Sahota: [00:44:06] And they're coming face to face with some of these really technical concepts, maybe feeling a bit discouraged and showing a lack of motivation. Do you have any tips to share with our listeners for how to stay motivated in their paths? Paul McLachlan: [00:44:22] I think motivation is very hard to answer general advice for. I think it probably goes down to what got you as an individual interested in this career path. And I think what I try to do is to always carve out time in my schedule in my week to do things that are kind of fun. So for me, that is reading about a domain. I don't know very much about or talking to people whose work I don't know very much about because I find that such a fuel for my own creativity. But what that might be, what that time might be, is going to be very different for each person. But I think oftentimes, particularly for people who are in school, we can be so heads down in terms of finishing a problem set, finishing a project, getting something done on a particular deadline that we forget to actually have fun. And particularly now, when there's so much uncertainty, I think people are giving them giving themselves even less time to have fun. I worry that that's a very easy path to burnout, to really bad imposter syndrome. And so whatever it means for you to have fun, I don't think that that is wasted time or just, you know, self gratification. It's actually incredibly critical for every person to stay motivated and to find creativity. So I think you have to actually think of having fun and staying connected and staying entertained is actually part of your job responsibilities rather than something that can be they can be set aside. Harpreet Sahota: [00:45:59] 100 percent agree with that. In order for you - Harpreet Sahota: [00:46:02] Learning is kind of state dependent, right. So if you are trying to learn something new, it's always good to operate out of the place of joy and inquisitiveness. Right. Like you mentioned, have fun. Infuse much of that as you can in your learning process. It just makes the process so much more enjoyable. Paul McLachlan: [00:46:19] I think that word, joy, is really it's a good one, because if, you know, if you are not finding joy or fun or entertainment or whatever word resonates with you and you and your work at any point during the week, you know, you have to find some way to build that in because it's critical to one longevity, because I think one thing that I'm still needing to learn myself is, is career and accomplishment. It's a marathon. It's not a sprint. So, you know, you can kind of push yourself for a couple of weeks or a couple of months, but that's not sustainable for 10 or 20 or 30 years. Harpreet Sahota: [00:47:03] Excellent point. So last question here before we jump into our lightning round. What's the one thing you want people to learn from your story? Paul McLachlan: [00:47:13] Oh, my gosh. That is a really good question. I would say, because I think I struggle a lot with impostor syndrome. And I especially struggled a lot with impostor syndrome, while I was in school. And I think the thing that I want people to remember is you don't know the story of the person who is sitting across from you or sitting at the desk next to you. Or is your T.A. or your professor. And we assume that everyone has had such a straight and linear path and everyone's everyone is gone from success and peak and peak and peak. And there's never been a valley. There's never been a setback. And that's just not true. Everyone has valleys. Everyone has setbacks. And it is just critical to keep in mind that everyone you're interacting with, from the CEO, to your classmate to your professor, is a human being. Just like you and I. Paul McLachlan: [00:48:15] I just keeping that in mind to me has really helped. It's not solved, but it's helped some of the imposter syndrome. And it also just creates an opportunity to connect with people at a human level. Because, again, we're all humans. And I find a lot of inspiration and a lot of creativity and learning people's story. Harpreet Sahota: [00:48:34] Very, very beautifully put. Harpreet Sahota: [00:48:35] I've got chills going on right now. That's really well put. Thank you for that. Harpreet Sahota: [00:48:40] So let's jump in the lightning round here. What's your Data science superpower? Paul McLachlan: [00:48:46] I guess I wouldn't call myself pretty creative. Paul McLachlan: [00:48:52] And that means to me, I. I really like finding strange or different ways to measure topics that people haven't figured out before. So I do a lot of creativity thinking. I do a lot of reading around different measurements, and that's probably my superpower. Harpreet Sahota: [00:49:10] I like that a lot. And I've heard you mentioned creativity throughout this interview, which I think is pretty awesome. What can a Data scientist do to cultivate creativity within themselves? Paul McLachlan: [00:49:22] I would say read things that are not things you're comfortable with. So read books, read papers, read newspaper articles about topics that you don't necessarily gravitate to because it's gonna open up a lot of other worlds. I also think that I will be honest and say I, I tend to find my best ideas come during physical exercise. So I know when I'm running or biking, I don't know what it is. Paul McLachlan: [00:49:47] I will be kind of maybe meditating while I'm doing that. And then an idea will come to me. So probably incorporating some physical activity is it is a great way to to just let your mind drift and think in a different way. Harpreet Sahota: [00:50:00] Yes. Very, very good point. Yeah, definitely. Anytime I'm on my walks or any time even working out in the morning. That is really where the best ideas happen. Like my morning is. Usually I do my cardio and then move right into a meditation. So going from the sympathetic to the parasympathetic nervous system, slowing everything down and rest and digest and just let those ideas bubble up to the surface. Harpreet Sahota: [00:50:27] So what is an academic topic outside of Data science that you think every Data scientist should spend some time researching? Paul McLachlan: [00:50:39] I mean, I think that every data scientists should have to take a social science class. I'm not saying that as someone who is trained as a social scientist, I think it's very critical because often in computer science or Data science programs, we work with synthetic Data or simulated Data. And we don't have to spend as much time thinking about these human processes that generate Data the sources of biases come up with that, the measurement error. They can come up in that. And then as soon as we're on the job, we suddenly get hit with a ton of human generated data. And it can be a really hard shift going from working with and building models that try to mimic efforts toward distribution. We know well to human generated Data. And I think that social science gives us really important tools to interrogate Data generating processes, to think about sources of bias. Because humans are not like machines, we can we can have all kinds of measurement error that comes up there. Paul McLachlan: [00:51:35] And social scientists have done an incredible work and building tools and methods to overcome those measurement error problems. Harpreet Sahota: [00:51:46] Thank you for that. Yeah, that's actually, you know, ask this question a lot during my interviews. And social science is the one that comes up most frequently. So there's definitely something to that. Yeah, definitely a few. Harpreet Sahota: [00:51:58] So what's the number one book of fiction or non-fiction or both that you'd recommend our audience read and your most impactful take away from it. Paul McLachlan: [00:52:09] The book that I think has shaped my thinking the most is Connected by James Fowler and Nicholas Christakis. So I think a lot you know, I work at a networking company. I think a lot about social networks. And I don't mean, you know, Facebook or Twitter or Instagram in this case. I mean the way that humans interact. And that actually has really interesting properties, both from a statistical sense in terms of how we think about statistical error and I.I.D. properties, but also in terms of how we think about nudges or contamination or spread of information. And I really think that this has kind of interesting implications for data science and society overall. But until reading this book in grad school, I really hadn't thought about how connected in networks we all are. Harpreet Sahota: [00:53:01] What's the weirdest question you've been asked in an interview? Paul McLachlan: [00:53:07] So I don't have a weirdest question, but I do have a weirdest process. And I actually think that there is a lesson here, which is when I first decided to leave academia, I applied pretty broadly and I had no idea what the job interview process looked like, what was normal, what was abnormal. And I really wanted to I really wanted to find a good job. So I didn't set boundaries. I didn't set limits. I just I kind of ran with everything that came up. So at one point, I had a phone screener interview that lasted for three and a half, nearly four hours. I think I got off the phone at midnight on a Friday night. And I realize now what the person is is really curious about me and what I could bring and had some questions around Data science myself. But I didn't realize that for our screener, interview is not normal. So I think that lesson for me here is to even if you're interviewing for your first job, it's really important to set boundaries for yourself. Harpreet Sahota: [00:54:12] What's your favorite question to ask during an interview? Paul McLachlan: [00:54:16] When can you start? That's a great one. It's mostly because in my interview. I'm really rooting for you. I want you to be part of the team. Harpreet Sahota: [00:54:24] So I - that's the favorite question. When when I can, you know, say everything went really well. Harpreet Sahota: [00:54:32] So if we could somehow get a magical telephone that allowed you to contact 20 year old Paul, what would you tell him? Kind of set the stage for us a little bit. Well, we're a 20 year old, Paul. At what are you doing? And you know, what advice or what would you tell him at that point? Paul McLachlan: [00:54:48] I was an undergrad at Columbia in New York. I was biking to and from campus every day. I would say probably the thing that I would love to tell myself is and it's a lesson I still need to learn. So this tells you how effective I am and telling myself this but that careers are Paul McLachlan: [00:55:12] Marathons, they're not sprints. Paul McLachlan: [00:55:13] And the difference between an A and an A minus is probably not so meaningful that you need to study an extra 10 or 20 hours or do all of the extra credit you possibly could. No one's going to check in fifteen. Here is, you know, how you did on chemistry 2. If you've got to be a plus or a minus. It seems really important and meaningful in the moment. But. It probably isn't in the grand scheme of things. And there's probably some lessons in there, I still need to learn. Harpreet Sahota: [00:55:45] So what's the best advice that you have ever received? Paul McLachlan: [00:55:50] Speak more slowly. Paul McLachlan: [00:55:53] That is, I. I have gotten some incredible mentorship from a lot of people and our marketing team and particularly around as a data scientist. Paul McLachlan: [00:56:03] Communicating effectively. When I get nervous, I try to I start speaking really, really quickly because I want to impart all of the knowledge I have and all of the excitement I have. So I really have to be mindful and speak more slowly and not actually allows people to engage with my ideas. And what I'm saying versus just trying to give it all to you at once. What motivates you? Puzzles. And that is solving puzzles. I look at some I can find something. And my my expectation is, is we should expect X, but instead we see why. When I can figure out why that happens, that that's the course thing. That's what I look for. Harpreet Sahota: [00:56:45] So what's the song that is giving you life right now? Paul McLachlan: [00:56:49] Little Dragon. Are you feeling sad? It's like three minutes. It's awesome. Go listen. Harpreet Sahota: [00:56:54] Oh yeah, definitely I'll check that out right after this. So, Paul, how could people connect with you? Where can they find you? Paul McLachlan: [00:57:01] I'm on LinkedIn, it's probably the best place to get me and Paul McLachlan. I work at Ericsson. I should pop up. And, you know, if you do have questions or want to reach out. I you know, just like you, I take mentorship really seriously. And I would be happy to answer questions if anyone would like to learn more about how to break into data science as a career or or anything else that we didn't cover in this conversation. Harpreet Sahota: [00:57:26] Paul, thank you so, so much for taking time out of your schedule to be here today. Really, really appreciate you coming onto the show and really enjoyed having you on and hearning your perspective on all of these things. Thank you so much. Paul McLachlan: [00:57:38] No, dude... Thank you so much. The honor is all mine. And it was an absolute pleasure.