The Case for the Deliberate Generalist

There is a quiet pressure in data science to specialise early. Pick a lane, the advice goes. Become the person who knows one thing deeply, because depth is what gets rewarded. I have spent the last few years doing something that looks, on paper, like the opposite. I have worked across seven technical domains, in three very different operating environments, on purpose. This is the case for why.

Breadth as a decision, not an accident

I treat government, research, and engineering as three parallel lines rather than three chapters that follow one another. In the same months I have built dashboards inside a regulator, worked on genomics pipelines for a medical research institute, and helped ship a mobile app as a startup co-founder. The work spans climate risk modelling at CSIRO, flow cytometry automation at WEHI, a mental health application at the University of Melbourne, strategic intelligence at Consumer and Business Services and at South Australia Police, and laboratory data work at CSL.

That list can read as scattered. I would argue it is the reverse. Each domain teaches a different way of being wrong, and the lessons compound. Government taught me to frame a question so a decision maker can act on it within a day. Research taught me to care about reproducibility long after the result lands. Engineering taught me that a model nobody can deploy is a model nobody will use. None of those lessons sit comfortably inside a single field.

The thread is method, not subject matter

People sometimes ask what holds it together. The honest answer is that the subject matter does not. The method does. I try to start from the problem rather than the tool. Before I reach for a model, I want to know what decision needs making and the smallest amount of data that would let someone make it well. That habit travels. It works the same way whether the question is about crime trends, commodity prices, or a clinician reading a patient dashboard.

This is what I mean by generalist by nature, specialist by discipline. The nature is a curiosity that refuses to stay in one box. The discipline is a consistent way of working that I apply everywhere: understand the problem, match the method to it, report the result honestly including its uncertainty, and make sure it can actually be used. A generalist without that discipline is just a dabbler. The discipline is what turns breadth into something an organisation can rely on.

Why this is a strategic choice now

The case for the specialist rests on a stable world, one where the deep skill you build today is still the scarce skill in five years. That world is getting harder to find. Tasks that used to take days of scripting can now be drafted in minutes. When the mechanical part of the work gets cheaper, the scarce part is no longer the leg work. It is the judgement about which question is worth asking, what level of risk is acceptable, and what a result means for the people it affects.

Judgement of that kind is built by seeing many situations, not one. A person who has framed problems inside a government compliance team, a research lab, and a startup has a wider library of patterns to draw on. That library is exactly what is hard to automate and hard to copy.

Where I am taking it

My aim is not to be the best at any single technique. There will always be someone deeper on a given algorithm, and I am genuinely glad they exist. My aim is to become the best data generalist I can be: someone who can walk into an unfamiliar problem, frame it from first principles, and select the right tools, knowledge, and people to resolve it. The seven domains are not a detour from that goal. They are how I am building towards it.

If you are early in your own career and feeling the pressure to narrow down, I would offer one gentle counterpoint. Depth and breadth are not opposites. Choose your breadth deliberately, hold yourself to a consistent method, and let the range itself become the specialism. That is the bet I am making, and so far it has been the right one.