Andrew Matte is a Technology Executive with expertise in data science, software engineering and artificial intelligence. Andrew is a language-agnostic technologist who strives to work on meaningful projects. His primary reason for learning to write software was to develop artificial intelligence applications and has found joy in developing several such production systems. Andrew Matte attended the University of Waterloo for mathematics and economics, the Canadian Securities Institute for several securities courses, and practices lifelong learning.
Connect with Andrew Matte on LinkedIn.
Resources mentioned in this episode:
- Lighthouse Labs - coding education
- hvr - social browser that makes it easy to discover and chat about the content you love
- TensorFlow - end-to-end open source platform for machine learning
- PyTorch - library for deep learning on irregular input data such as graphs, point clouds, and manifolds)
- Kaggle - world's largest data science community with powerful tools and resources to help you achieve your data science goals)
- Alex Rascanu’s LinkedIn profile - career development advice
Transcript of the interview:
Alex Rascanu: Welcome to The Career Planning Show. Our guest today is Andrew Matte. Andrew, how are you?
Andrew Matte: I'm doing great. Thanks. How are you today?
Alex Rascanu: I'm very well. Thank you. Andrew, can you walk us through your career journey?
Andrew Matte: Sure. I worked at various places when I was a kid and a teenager and then, in my undergraduate years, while I was in school I worked in what you could call customer service or hospitality as a waiter in a restaurant for three years. But my summers I spent at Friedberg Mercantile Group interning as a research analyst. I studied math and economics at Waterloo [University] for a number of years before dropping out to travel. I got great advice from my dad who said, "Andrew, it doesn't seem like you enjoy being a student very much. You are learning. We can tell that you do a lot of studying, but why don't you take a little bit of time off?" And so I did, and I went out to Whistler, British Columbia during the 2010 Vancouver Olympics. For those of you who don't know, Whistler also hosted some of the events, being just a few hours North of Vancouver. I got the job thanks to my languages. It was another hospitality- type role where , at Canada Post, I handled the cash register. But the Olympics ended and so did my time at Canada Post. The work ran dry and the economy of Whistler really shrunk after that event dissipated. I came back to Toronto, Ontario, where I grew up and I worked back at Friedberg Mercantile Group as an investment advisor. It was a sales position.
There are a few rules to working in finance. One of them is KYP, know your product. So I had to do a lot of research. KYC, know your client, which means that you have to observe fiduciary duty in order to make sure that you're putting your client in the best investments for them specifically. There are other few principles around that, but we'll stick with those two for now.
I was there for two years. I learned a lot. I learned that doing research and studying economics and math doesn't prepare you for a sales job. I really did enjoy the research side of things. Eventually I moved on to some research positions as a data analyst and, thanks to my programming expertise , I was able to make dashboards. I took an SQL course, which stands for structured query language in order to pull data directly from relational databases. And one thing led to another.
After a few years of picking up better practices at Lighthouse Labs, I moved into more of a leadership role as the head of technology at HVR Technologies, which is still in development. It may be released soon. I can't say too much, but it's a social browser. It's a social network right in a mobile browser for both Android and iOS.
Alex Rascanu: You're engaged with some really interesting initiatives at the executive level within the technology space . If we go back to two items that I picked up in your career journey, one is languages and the other was your family's point of view with regards to what you could be doing with your life, which is "maybe take a year off." In terms of languages, I guess you are not talking about programming languages that came in handy during the role in BC. Can you tell us what those languages are and how did you manage to pick up new languages?
Alex Rascanu: Being able to pick up new languages is a skill that can really come in handy throughout our lives on a personal and professional level as well.
Looking back at your parents' advice that you take a year off from your educational experience. You were going to a top school , Waterloo, you were in a good program that was playing to your skills... Ultimately, you are still using math and economics in various ways in the work that you're doing in the technology space... Looking back, what do you think about that advice and how has that advice from your family impacted your, your trajectory?
Andrew Matte: You know, I look back at that advice and I think it was the right advice. I did get four years of education in and with the unfortunate thing is that employers don't really care about education. In most cases, if you're going into a regulator profession, it's vital. But software is largely unregulated unless it's being applied to regulated industries. Most industries are, I would say, largely unregulated in terms of whether or not you need to have a degree . What are the barriers to entry to say running a successful podcast? It's just competency and you have to have interesting guests. So not only looking back at that decision at that very moment, but in the context of my career, employers have largely asked "what can you do? What have you done?" And, in many cases, like my interest in artificial intelligence and data science, I had to prove it out through personal projects and portfolios before I was able to equivalently ask for money for it at the job.
Alex Rascanu: That's great. Let's shift over to a term that you use, which is artificial intelligence. I know that's a space that you've been in for the last few years. Can you explain what artificial intelligence means?
Alex Rascanu: That's right. Then, from a revenue perspective, anybody who works in a marketing or a sales role, understanding that having a strong presence in Google search results is really critical. Can you talk a little bit about a phrase that you used recently when we were chatting offline , around how hungry search engines are for data and how artificial intelligence plays a role in capturing any new content that gets featured online?
Andrew Matte: Search engines make money by showing relevant content, alongside ads that are also hopefully relevant. And in order to show data, they have to capture that data and metadata and digest or index it in a database so that it can be easily recovered and surfaced when the appropriate query shows up. In the context of ensuring your presence online, to show up in Google search, they say "content is king." By surfacing or publishing regular data about your core competencies, your customer's problems and your solutions for them...
Alex Rascanu: Make sure you have ongoing blog posts going live, video content...
Andrew Matte: Yeah. And more than just video content is also indexing that video content properly, maybe posting a transcript alongside of it. Most of the search engines out there parse text very well. So what that means is that if your content is in text , rather than audio files or video files, then you can really make sure that the search engine that is crawling your site in order to display the results is able to parse the text and understand algorithmically the context in order to surface it at the right time.
Alex Rascanu: That's really helpful. In terms of the technology executive function that you find yourself in in the last number of years, can you speak to what that entitles? What does it mean to be a technology executive practically on a day-to-day basis?
Andrew Matte: On a day-to-day basis, I would say at the heart of it is competency in the technology that you are an executive of. In my case, that's web technologies and machine learning or artificial intelligence. There's also of course, the leadership component and understanding the future of the technology in the business. I really believe that technology should inform and follow the business rather than be the business. Otherwise... there's the analogy of a hammer and a nail. If all you have is a hammer, every problem is going to look like a nail. So you really want to understand what the company's core competencies and mission statement are in order to make sure that technology is facilitating and augmenting that.
Alex Rascanu: If we take a step further from there and we think about how technology can be used to positively impact society... I know that this is something that you're passionate about. Even the project that you're currently working on is around trying to parse significant amounts of data and find ways to improve services that the citizens. Using technology to parse data at such a fast rate that you're able to provide informed advice much more quickly than if someone was doing it manually is something that's highly beneficial. Can you talk in broad strokes as to how you perceive that the use of technology can have a positive impact from your experience?
Andrew Matte: Right. So I have a framework that I use, which is about data and, and tasks around data, which is to collect, analyze, augment, and automate. So it's a four-step process. If you want to get to automating a specific task you have to develop some kind of software that will collect data about that task. And then sometimes there's a back and forth between analyze and collect. If you don't understand the problem space fully, you're probably going to go from analysis back to collecting data a number of times in order to ensure that you're getting all of the data that you need to really be able to predict what comes next. Then the next stage is augmentation, which is when you're able to integrate your findings into the process in a way that will improve the decisions and output of the participants who are actually carrying out the task. An example of that would be if you're an auditor of, let's say mortgage applications, and you're not necessarily an underwriter, so you don't have all the knowledge or skills involved in gaging the risk on an incoming loan application, but it's already been funded. And you need some kind of tool to get a barometer reading of how much risk there is on this loan to prioritize a large number of loans in order to pick out which ones to audit first. This is an actual application that I worked on at Paradigm Quest here in Toronto, it's a mortgage servicing agency. We collected data from the database, thanks to the application, and we analyzed the data and saw that we had a lot of data around whether or not an underwriter said yes or no to a mortgage application. We analyzed it and then we eventually came to a process where we were able to augment the auditor's experience by providing that barometer reading. And we never automated it. We never said these applications are going to be declined or approved based on a computer, but we did get to the stage where the auditors were very happy with the tooling that they had so that they could prioritize which mortgages to examine first to audit and to really dig into. And sometimes it was just a matter of, well, the I's weren't dotted, the T's weren't crossed, but there was at least one instance of a loan being funded when it shouldn't have been and measures were taken in order to correct and remedy the situation. That would be an example of collect. Analyze augment and not quite automate, but it's a good example to understand how you can augment an audit process. Alex Rascanu: That's great. Now if we find ourselves taking a look at the global environment and the kind of challenges that we see in society , I remember us recently talking about the employment services space /the career planning space and looking at the United Nations sustainable development goals in the area of workforce development. I'm looking at a few other areas where you find government entities, you find non-profit organizations, you find corporate entities really engaged to try and find solutions to some significant societal problems where if we address them correctly through different policies and programs and the coming together of the various stakeholders, we can benefit the significant number of people. I know that these are some of the things that you think about. Would you mind speaking about how the use of technology, whether it's artificial intelligence or some of the other tools that you have in your toolbox, how they can be leveraged to be able to identify the solutions and potentially the ways in which those solutions can be executed? Andrew Matte: Yeah. In order to leverage artificial intelligence in any solution, you have to understand what artificial intelligence can do and what it can't do. And we're currently at a stage in artificial intelligence research, from the academic papers being published to the open source, frameworks being put out like a TensorFlow and PYTorch and so on... those frameworks and tools are what you would call specific artificial intelligence. What they do is they take optimization on a single task and they find basically a math equation using programming that will repeat the prediction and -I say repeat loosely here, because it's not exactly repeat, as there are times when you can use it to come up with novel solutions or solutions that haven't been seen yet - the further you go out into that space and the more generative you are with your solutions, the more quality assurance I would say is necessary from a human who understands the space and the impact of the decision. Going back to how a company can use artificial intelligence to improve what they do, is within that framework of collect, analyze, augment, and automate. Understand what your final step is in automation. And really, I think the right decision is not to replace people, but to make their jobs better and easier and faster so that they enjoy their time more, so that they don't have the same cognitive burden when they're performing tasks. And a lot of that comes down to getting a good data set. So if you have a problem that you want to automate, you have to collect data about people who are actually doing that thing. Collect the data; that's the first step.
Alex Rascanu: That's great. If someone is inspired by some of the things that we're talking about and they're hoping that one day they could become a technology executive, are there any thoughts that you have about the process that someone could follow in terms of their educational and work experiences?
Andrew Matte: Yeah, absolutely. The first thing is core competencies. And in terms of artificial intelligence, a great -let's call it playground- is Kaggle K a G G L E, which was acquired by Google . They call themselves the home of data science and they have a funny blog slogan, which is "there's no such thing as a free hunch," which is a play on there's no such thing as a free lunch from economics. They have tutorials. They have notebooks of people publishing open source analysis, exploratory data analysis, and predictive algorithms. They are a great place to keep aware of what's going on in the space because it's an open source competition model. And so they'll put out a company's data with cooperation from the company, and they'll say "this is what we're trying to predict. This is the metric we're using in order to prove to assess a success. And, given that, this is when the competition starts, that's when the data will be available. This is when the competition ends." You can have one to, let's say, five submissions a day. Some of the competitions are open submission, so everybody gets to see your code. They'll execute it for you. And other competitions are closed code format so you'll only have to release your code to the company and only if you win. One of the trends I've noticed was at first there was a wild diversity in the kinds of solutions. Back in 2014, when I started doing Kaggle competitions solutions ranged very wildly, which algorithms people used and whether or not they leveraged what's called transfer learning and pre-existing models in order to understand the space better. At first Kaggle open sourced the notebooks and they called them kernels at the time so that you could share your work with other people and give people a head start on the competition is a great way to open source. They introduced prizes for that for more useful Kaggle kernels and open-source notebooks. In the past few years, I've noticed that there's an incredible amount of transfer learning that is being applied. The competition is getting... I don't like to use the term arms race too much, but the bigger computer you have, the more resources you have available, the more likely you're going to do better than the competition. If you have access to models that were already trained, you're going to do better.
Alex Rascanu: That's fascinating. So you see that as a training ground.
Alex Rascanu: That's great. Andrew, thank you so much for all the insights that you've shared on this episode. We very much appreciate it.
Andrew Matte: It's my pleasure. Thanks for having me.