From Code to Data: Mike Gouline on Leading Data & ML at mx51

We chat with Mike Gouline, Head of Data & ML at mx51, about his career journey. From his early days in software and mobile development to leading innovative data engineering and machine learning projects, Mike shares his insights on the evolving landscape of data roles and the future of the field in 2024.

Please introduce yourself

Hey, my name is Mike. I live in Sydney and I’m almost certainly not an AI.

What do you do for work today?

I work at an Australian startup called mx51 where we bring innovative merchant experiences to our banking customers. My current title is Head of Data & ML, which means I lead a small team building our data platform that powers business intelligence and embedded analytics.

What did you study?

I studied computer science and focused on machine learning and theory. Unfortunately, machine learning was only done in academic research around here in 2010, so my dissertation is where I left it and got myself a coding job.

Can you describe your career journey to date?

My earliest jobs were miscellaneous web development, but a geeky obsession with mobile phones soon landed me in Android and iOS right as everyone bought their first smartphones and companies needed apps… a lot of them! Over the next 8 years, I worked on mobile projects for companies like Foxtel, Google, PwC, CoinJar and Cochlear. Much of the excitement over seeing your creations on smaller screens eventually wore off and I started tinkering with Python and Go, before quitting for a role that unexpectedly transitioned to data.

What was your experience moving from mobile and software development to data engineering?

As much as I wanted to return to machine learning, my move was out of necessity. Working on backend services at mx51, we needed a way to run analytics across multiple databases and soon enough that became my main responsibility. Messy scripts gave way to a deliberate architecture that I was so determined to perfect; I would excitedly run to my whiteboard whenever an idea struck in the middle of a lockdown walk. Once basic functionality was stable, I could hire a team and start doing more interesting things with that data.

What does the role of a data engineer look like in 2024?

Data engineering means different things to different people. Anything processing data to answer analytical questions falls under that term, however, what that looks like and how broad an engineer’s responsibilities are depends on the company.

Startups favour managed services, allowing engineers to have broader influence. Using off-the-shelf products like Snowflake, Databricks and Fivetran gets you pretty far without much code, other than instrumentation to make them all work together. Like serverless in web development, it frees engineers from having to maintain infrastructure to focus more on business logic.

By contrast, larger organisations might have more complex architectures where it’s impractical for one engineer to deeply understand the whole system. Entire teams often just maintain Hadoop clusters. While the industry is slowly moving away from this, enterprises are reluctant to retire perfectly good production systems and even modern services like Databricks can rely on Spark.

Most companies sit somewhere in between these extremes, but a data engineer’s mission is ultimately turning raw data into useful insights, by maintaining infrastructure, writing queries, training machine learning models, or whatever else.

What does a typical Data team look like? Where do they fit?

What your employer expects from data determines whether it’s seen as an engineering discipline, a role within any team, or both.

Data as an engineering discipline looks no different to backend or mobile, except you work on components that deliver analytics. The proximity to other engineers tends to bring more of the software development process into it and often attracts more technical people. Whenever research is a large component, data teams can also be R&D.

Where data is seen as a business process, rather than a product, data analysts are generally implanted in the teams requiring analytics, such as finance or operations. The orientation shifts from engineering practices to outcomes, so you see more skills on the visualisation end of the spectrum.

A hybrid approach appears whenever a centralised data team lacks business context or dispersed data analysts need a common infrastructure. This normally applies to companies with larger data contingent.

What advice would you give to someone interested in pursuing data engineering roles?

Data engineering is fragmented, pick a corner and get learning! You may not have experience with petabytes of data, but online resources and free trials of cloud services can get you far.

What are some aspects of data engineering jobs that might surprise people?

The outsider perspective is a superpower here. I’ve met data people who came from academia, software engineering, and even accounting, and they all bring insights that even seasoned data professionals lack.

When hiring a data professional, how do you assess the strength of the candidate?

You won’t find anyone with experience in your exact data stack, why assess candidates on specific tools? My questions revolve around what they already built and why, or how they would architect a hypothetical requirement. Another trick is picking a technology from their resume and asking what they hate about it – no matter how much you love Python, you have horror stories and I want to hear them!

What is your favourite tool or resource, and why?

I’m known for my weird fascination with DuckDB. Yet to use it for anything in production, I love analysing local data with it and generally hold them as an example of what open source should be.

Any podcast you are currently listening to, or a book you are reading?

Decoder and CoRecursive are awesome, and my unoriginal data book recommendation is “Designing Data-Intensive Applications” by Martin Kleppmann.

Toughest work moment?

“Back to the drawing board” moments. They happen a lot in data engineering where you design for a particular scale and surpassing that often requires a way bigger fix than bumping up your instance.

Most rewarding work moment?

Seeing somebody use your work, whether it’s a colleague debugging their issue with your data tool or a cochlear implant recipient controlling their volume with your app.

Your one-sentence work-related advice

Technology is evolving, think about where you want to be next and then figure out how to get there.

Your one-sentence hiring-related advice

Optimise for learning capacity and perspective, not years of experience in a tool released last month.

Check out more from Mike here. Find him on LinkedIn here.

Subscribe to our monthly newsletter to get our latest content in your inbox.