Behind the data scientist
Welcome back to
Behind the data scientist
Hello everyone, welcome again to this series of Data Scientist interviews by Shapelets!
As you know, we will be interviewing experts in the Data Science field to explore different topics in the data science community and put the Data Scientist professional into the spotlight.
Today, we are very happy to be joined by Matthew Kirk, Principal Machine Learning Scientist at Zeitworks. He has extensive professional experience in delivering value-added AI/ML solutions.
In this interview, we discuss the main challenges data scientists face when providing solutions and insights to third parties. We also highlight that including the data science community in efforts is essential, and also the importance of their data science contributions. Don’t miss it and have a look at the video below!
Watch the interview here below, don’t miss it!
Could you please share your professional background with us
Matthew Kirk – So, I started my career as a Quantitative Analyst in Finance. Worked at Stanley eventually, and I did really like finance that much and found myself transitioning into startups. So, I worked at startups for the past 11-ish years. My background is really mathematics and statistics, but also engineering. I kind of mixed the two of them and become the data scientist, machine learning type of person. That is what I’ve been doing for the last 11, 12 years.
Why did you choose Data Science?
Matthew Kirk – You know, Data science is interesting cause I was writing software, for the most part. You know, be on websites, and someone asked me “I know you have a background in mathematics, so could you do this other project?”. This was around 2009, 2010-ish, and then I quickly found myself going down the road of machines, clustering, … I kind of got pulled into it through the problems we were facing. I’ve just kind of been following that ever since. It wasn’t like a thing when I say to myself “I’m going to become Data Scientist”. I just kind of found myself on the wave pretty early.
What do you like the most about your job?
Matthew Kirk – The thing I like the most is really the learning aspect. You know, I went to school, I did my Masters’s degree and that’s great. But I find that in Data Science, Engineering, and Software I’m just constantly having to reinvent myself. The technology that I use today, is so different from the technology I used 10 years ago. And I really enjoy that, I enjoy learning what’s happening in the community. I also really enjoy learning from other people. So I find that the collaboration and the social aspect is a lot of fun to me and that’s what I really like.
Who is your role model?
M.K. – You know, this is a hard question. There are a lot of people I look up to in the community. But I think the role model for me, is my dad. My dad is a huge role model for me because he was not college-educated, and he was super smart. And I find that I embody a lot of what he did. I wake up early in the morning, go get that attitude, I think it’s really important, especially data scientists, we just need to keep going, and keep trying, just that positive attitude.
What advice would you give to data scientists?
M.K. – I get asked a variation of this question in workshops and events. And I think it’s so personal. What I would recommend is to really understand your own personality, and yourself. Like, “What are your strengths and weaknesses?”, and get really clear. So, when it comes to data science, I really struggle with some of the higher math. I find that the really complicated mathematics is tough so I tend to focus more on learning that stuff. But other people would struggle with programming, so they have a hard time writing Python, so have a hard time with that. It really depends on the person. So I very much recommend the approach of taking programming, mathematics, statistics, and communication, and really ranking yourself as to “Am I good at those?” and from there, start to work to improve yourself. So work on the weaknesses and lean on the strengths.
We are all different, we all have different gifts to bring to the world, and strengths and weaknesses. And I think it is better to know that.
What are the main challenges for data scientists nowadays?
M.K. – The biggest challenge I see right now, something about the 80% of the projects do not get into production. The biggest challenge in Machine Learning or Data Science is shipping it. There is a bit of community right now, which I think is starting to get there. But it is a huge challenge. My recommendation is to learn cloud computing, operations, scaling… It is hard but it’s so important. Because really, we are not done as data scientists until we impact the user.
What do you value the most in a data analysis technology or software?
M.K. – I find that when I am looking at new software to adopt, I really appreciate companies or organizations that build a good community. So, it is interesting that big companies have great tools, but open-source and building communities are really important. Because you may have the best piece of technology in the world, but if nobody uses it, it is like nobody cares. I look at the community and what people are saying about new technology. Some companies are very inclusive and will listen to the people. And some others are like “You should love the work we do”. So some companies that are really hard to work with and I think is just like an attitude problem.
What approach do you follow on a data analysis-based project?
M.K. – There are a couple of approaches to flow around a project. There is CRISP-Dm that people use. I tend to follow a simpler version which is TACT: target, arrange or acquire data, compose, and transmit like ship it. That is the approach that I take. I find that getting the goal first is so important for data science because, otherwise, it’s just experimentation, which is good too but in a different way. I think it is better to have a frame of what are we going to do and follow from there.
What skills does a data scientist need for 2022?
M.K. – I think that the two main skills are, first operations are really important to learn it. Understanding operations, it could be local or not, just understanding it enough. The second thing I would say is communication skills. Being able to communicate effectively is huge in data science. Even if you have the best model in the world and nobody is paying attention to it, it is kind of like it was all for none. I would say that in other data science-specific things if someone is trying to stretch themselves into new arenas, I think that reinforcement learning would be the first thing that will probably take off. This is my prediction. So, Torch and PyTorch in reinforcement learning is a very interesting project. So I’m focusing on these things right now…
Digital Marketing Specialist
Say hello to our Digital Marketing Specialist! Fátima’s role at Shapelets is to plan and execute digital marketing strategies and content to creatively develop and optimize our business on different platforms. She specializes in SEO, social media and digital content.