A day in the life of James Mylroie-Smith, Head of Data Science at Datium
A day in the life of is a series that champions Tech voices in the APAC region, to inspire anyone looking to begin a career and/or pivot into Tech.
TL;DR | This is for you if you’re
- Looking to get advice on starting out as a Data Scientist in the APAC region
- Wondering whether you should start your Tech career in a startup or a bigger company
- Figuring out whether the companies you’ve shortlisted really “do AI”
Rebekah: Hey James! Thank you so much for taking the time to speak with SUPA. Let’s kick off with your day to day and what you’re up to now at Datium.
James: Thanks, Rebekah. I’m currently the Head of Data Science & AI at Pickles/Datium where I manage a team of data scientists. We’re an auction market place and Datium provides products to leverage data from Pickles.”
Rebekah: Interesting! What type of problems do you work on at Datium?
James: We look at asset pricing mostly. We want to be able to value these consistently and fairly in the ever changing market. We’re always trying to find business value and looking at ways we can automate processes that have a lot of domain knowledge.
Rebekah: How do you align with stakeholders on what you do vs what the business’ goals are?
James: We have a lot of questions and conversations with key stakeholders in the business. Data scientists try to convert their goals to internal metrics. So we ask things like: “What does this metric mean for our model?” We then need to find metrics that we can talk to them about, in a way that makes sense to them. When we’re modelling, it’s not always a 1:1 parallel with business needs. They want to know where the business value is to better quantify the product for them.
Rebekah: That’s really cool. Was this combination of data and business something that interested you all along?
James: I actually studied physics at university. I went for physics because I found it quite interesting — there were exciting things happening in particle physics at the time! But it was then that I realized I enjoyed a lot of the data analysis. To me, the physics was interesting but there was more enjoyment from the data analytics, machine learning side of things. So I felt that was the angle I wanted to pursue and data science gave a lot more variety than physics.
To me, the physics was interesting but there was more enjoyment from the data analytics, machine learning side of things.
Rebekah: I love how you found your passion in data analysis and took that leap of faith to move into a new industry. What was your first data science role like?
Starting as a solo data scientist in a Tech startup
James: I joined a small start up as the only data scientist. So I was exposed to both back end engineering and data science work. It was very useful to get to know data in a different industry and being able to think about how machine learning would solve different kinds of problems.
Rebekah: How did you feel being the only data scientist in the company?
James: I had friends who went down a similar route as me, so I had people I could talk to — that definitely helped. For people just starting out, it would be useful to have a senior or a mentor to guide you to approach problems. It doesn’t have to mean you need to join a company with a big data science team, it could be a small one. Your mentorship could also come from a CTO.
Rebekah: If you were to go back in time, would you still have chosen the path of being the solo data scientist in a Tech startup?
Start ups vs larger companies
James: I would go back and take the same route. Startups give you a much broader exposure to everything. I learned a lot about things outside of my job scope like product building, data structures and databases. At a larger company, you will have a smaller area of coverage. I think both are useful experiences — whether you choose to go for depth or breadth. If you’re a fresh grad with a bachelor’s degree or masters, probably getting specialism might be a bit of an easier way to start.
Startups give you a much broader exposure to everything. I learned a lot about things outside of my job scope like product building, data structures and databases.
Rebekah: What degree programs or specialisations would you recommend to people looking to work in the data science field?
James: I would consider looking at data science degrees in general. But for some people, you might want to consider engineering if you like computing and then specialise in machine learning. If you’re more interested in business, then look for data modules in that. So basically, look at the electives available in the final years of the programme. Find out if you will be doing machine learning. You might also want to look at the pre-requisites. Find out if the course is data-driven business or is it more business management? So it’s important to look at where your interest in data is coming from when picking your degree.
Finding the right company in AI
Rebekah: Let’s say I’ve already picked my degree and I’m looking for my first role as a data scientist or machine learning engineer. How do I find a place that would suit my interest in data? I know a lot of companies say “they do AI” but the reality is a lot different!
James: James: Yes, for sure. Many companies don’t really do AI. There are companies where data scientists don’t do machine learning — they use spreadsheets and hold a more business intelligence kind of role. I would suggest asking a lot of data-related questions during your interviews. For example, what’s the data maturity in the organisation? How do they consume and use data? What data do they have and what volumes do they work with? If the company doesn’t have much data, you might struggle to use AI/ML. So, clarify what your day to day might look like and understand the data duties that’s expected of you in the role.
I would suggest asking a lot of data-related questions during your interviews. For example, what’s the data maturity in the organisation? How do they consume and use data? What data do they have and what volumes do they work with? If the company doesn’t have much data, you might struggle to use AI/ML.
Rebekah: Speaking to day to day, what does your typical day look like?
James’ day as a Data Scientist
James: A typical day would usually start with spending a bit of time testing out ideas, maybe training a model or testing out an approach. And then usually about half an hour to an hour of calls with the team — more checking in and team discussions to see where people are having problems. Then after lunch, I’ll have 1:1s with team members, kind of deep diving into problems they’re having — could be in work or be related to career growth. Aside from that I also have to update stakeholders with our progress, what’s en route to production, where we are in the process and getting feedback.
Data labeling in data science
Rebekah: Nice! What role does data labeling play in the life of a data scientist?
James: Generally, early on in a problem when you’re first doing a proof of concept (POC) for a task, it’s a bit heavier. You do a little bit more weak labelling of inspecting or tagging images. Usually that would be a few hours a day if required — to get a baseline to start understanding a problem.
Rebekah: So you mention a few hours a day of labelling. Wow! What are some assumptions that people make about the day to day of a data scientist?
James: A lot of the time people think it’s machine learning. A lot of junior/entry-level data scientists want to spend a lot of their time doing machine learning. But actually there’s a lot of work behind the requirements and getting the data right first.
A lot of junior/entry-level data scientists want to spend a lot of their time doing machine learning. But actually there’s a lot of work behind the requirements and getting the data right first.
Rebekah: What involves getting the data right?
James: There’s basically the cleaning of data, some element of inspection — checking through rows of data and making it sensible. Some preprocessing steps to make it helpful. Then, labelling of data. In that case, depending on the problem, if you’ve got an initial dataset you might not need to label. You can try and model it and see how it performs.
Pivoting from data science in Tech
Rebekah: Interesting! So it’s a mixed variety of tasks and not necessarily machine learning every day. What happens if after I become a data scientist and I realise this isn’t for me?
James: I’ve seen people pivot to business intelligence, data analysis and data engineering. Now there’s machine learning engineering as well. There’s also areas like Growth or even Marketing, depending on the skills you’ve learned, you could potentially integrate that into a Marketing role where you’re doing the optimisation of the Marketing funnel. You’re doing the data linked work, using the knowledge you’ve learned from data science to optimise funnels. Similar for Sales. Any element of the company that’s data driven has potential for you to pivot into.
Rebekah: What advice would you have for people looking to pivot into data science here in APAC?
James: Spend some time understanding a little bit about what is interesting to you. I think there’s so much variety of data science work. So that’d be the first part try to understand what you’re looking for. Also, talk to people about what you’re thinking about. Try to find people in your network and ask them what their day to day looks like, because that is what you’re going to be doing.
Rebekah: If there’s nobody in my network involved in data science, where should I go?
James: For Malaysia, there’s an active TensorFlow user group. I’ve seen some interesting hackathons that pop up here too. And that can be a nice way to build a network of connections with similar interests.
Rebekah: Nice. Finally before we end, how would you sum up what a data scientist does in one sentence?
James: We leverage AI and ML to add value to business.
Rebekah: Thank you so much for your time, James!
James: Thank you!
