MSc Data Science

Data science sits where statistics, computing and a subject domain meet, and the degree is honest about needing all three. A good programme teaches statistical inference properly, not just model-fitting; covers the engineering of data at scale, not just notebooks; and forces students to work with messy real datasets where the hard part is framing the question.

The market has matured: employers now distinguish sharply between analysts, data engineers and machine learning engineers. Programmes differ in which of those they actually produce, and reading the module list against target job descriptions is the most useful hour an applicant can spend.

The most quantitative of the computing master's, and the one where a mismatch between the applicant and the programme shows up fastest.

• A bachelor's degree in computer science, mathematics, statistics, engineering, economics or another quantitative field.
• Statistics and linear algebra at undergraduate level — genuinely required, and where most refusals originate.
• Programming ability, usually Python or R. Some programmes teach it; most assume the basics.
• English proficiency evidence.
• A statement showing you understand data science as statistics plus engineering rather than as a job title.

Programmes with this name range from statistics degrees with coding to engineering degrees with some statistics. The module list tells you which, and the two lead to different jobs.

Demand is real across sectors, though the roles under this heading vary widely in how technical they actually are.

• Data scientist — modelling and analysis, in tech, finance, retail, health and government.
• Data analyst and business intelligence, the largest volume of roles and a common and sensible entry point.
• Data engineer — pipelines and platforms, currently in shorter supply than data scientists.
• Machine learning engineer, more software engineering than statistics.
• Quantitative analysis in finance and insurance.
• Research and public policy analysis.

Job titles in this field are unreliable — two "data scientist" roles can be almost unrelated work. Read the actual responsibilities before deciding whether a role uses what you have learned.

One year across three strands — statistics, programming and the engineering that gets a model into use — plus a project.

• Statistical modelling and inference, the foundation everything else rests on.
• Machine learning — supervised, unsupervised, and enough theory to know when a model is lying to you.
• Programming for data — Python, SQL, and the standard scientific libraries.
• Data engineering — pipelines, warehousing and working with data at a size that does not fit in memory.
• Visualisation and communication, which is what most of the job turns out to be.
• A capstone project on a real dataset, often supplied by an employer.

Deep learning is the most talked-about module and rarely the most used at work. Regression, cleaning and communicating a result honestly are what most data science jobs consist of.