Data Scientist
You build models that predict or classify something valuable, and you own whether they hold up in production.
01
What the job actually involves
- Framing a business problem as something a model can actually answer, which is most of the difficulty.
- Building, validating and deploying models — and monitoring them once real data starts drifting.
- Explaining to non-technical stakeholders what the model does, what it cannot do, and where it will fail.
02
What you would need to study
- Statistics, Computer Science, Mathematics, or Engineering, usually with a Master's
- Strong mathematical foundation matters more here than in analytics
- A serious portfolio can substitute for the degree, but less often than in data analysis
03
Skills that matter
And which part of your Growth Profile each one builds, so they can be scheduled rather than just listed.
Python and its ML stack
Skills
Statistics and probability, properly understood
Academic
SQL and data engineering fundamentals
Skills
Communicating uncertainty honestly
Skills
04
What the work is really like
- Project-based, with long stretches of experimentation that produce nothing usable. That is normal, not failure.
- Close collaboration with engineering — increasingly you are expected to ship, not hand over a notebook.
- Generally good hours and strong remote availability.
05
The hard parts
Every career page here carries this section. A page that only sells is no use to you.
- The gap between a course project and production machine learning is very large, and most candidates underestimate it.
- Many advertised roles are actually analytics roles with a fashionable title. Read the responsibilities, not the heading.
- Genuinely entry-level data science jobs are rarer than the course marketing suggests — many people arrive via analytics or engineering.
06
Where it is heading
- Real and growing demand, concentrated in larger companies with actual data maturity.
- The field is shifting toward applied ML engineering and away from standalone modelling.
- Strong compensation, with a widening gap between competent and excellent.
07
What to learn first
- Mathematics for machine learning — linear algebra, probability, optimisation
- A hands-on ML course that requires deployment, not just a notebook
- MLOps fundamentals
- A domain: finance, healthcare, retail. Context beats another algorithm.
08
Projects that prove you can do it
These build the Experience dimension — the part of a profile you cannot revise for the night before.
- Deploy one model as a working service that a stranger can use. One deployed project beats five notebooks.
- Enter a competition, then write up why your approach lost to the winner.
- Reproduce a published paper's result and document exactly where you could not.
09
People worth talking to
Roles to seek out, not names. One honest conversation beats ten articles.
- A working data scientist about how much of their week is modelling versus everything else
- An ML engineer, about the handover problem
- Someone who went into analytics first and moved across
Knowing the job is half of it. Knowing whether it suits you is the other half.
The assessment ranks all ten directions against how you actually like to work, then builds the plan to get you there.
Free to start · Nothing shared without your say-so