Computational Scientist

A computational scientist uses computer simulation, modelling and numerical methods to answer scientific or engineering questions that would be too slow, expensive or impossible to answer by experiment alone. Typical work includes writing and running code that models physical, chemical or biological systems — anything from how a drug molecule binds to a protein, to how air flows over a wing, to how a weather system develops — and then analysing the resulting data. They work for pharmaceutical and biotech companies, engineering and energy firms, government research labs, universities, national computing facilities and specialist software companies.

Starting salary: £30k–£40k
Live graduate jobs: 2 open now
Main way in: PhD route
They usually ask for: A numerate degree is essentially non-negotiable

Approximate graduate salary

Roughly £30,000–£40,000 to start for graduate and MSc-level entrants, with PhD-level entry typically starting higher, often in the region of £38,000–£50,000. These are broad approximations only and vary considerably by sector (commercial pharma, tech and finance-adjacent roles pay more than universities and public research institutes), by location, and by how much the role leans towards software engineering.

What you'd actually do

Writing, debugging and extending scientific code — commonly Python for analysis and workflow, with C++, Fortran or CUDA (a programming toolkit for running code on graphics cards) underneath where speed matters.
Setting up and launching simulation jobs on a high-performance computing (HPC) cluster — a large shared machine where you submit work to a queue rather than running it on your laptop — and babysitting long runs that may take hours or days.
Analysing the output: processing large numeric datasets, plotting results, checking whether a simulation has converged or produced physically plausible numbers, and tracking down the cause when it hasn't.
Reading papers or technical documentation to understand the method behind a model, then implementing or adapting it.

Show 3 more

Meeting the people who asked the question — experimental chemists, engineers, clinicians, product teams — to translate a fuzzy scientific problem into something a model can actually answer, and to explain what the results do and don't show.
Writing up findings in reports, slide decks or internal wikis, and in some settings co-authoring papers or conference posters.
Housekeeping that quietly eats time: version control, automating repetitive pipelines, writing tests, managing data storage, and making someone else's undocumented code run again.

How graduates get in

PhD route — the most common entry point in research-heavy settings (pharma modelling, national labs, climate and materials science). A PhD in a computational or simulation-heavy area of physics, chemistry, biology, engineering or maths is often treated as the standard qualification rather than a bonus.
Direct entry from a numerate undergraduate or MSc degree, particularly where the work is more software and data engineering than novel method development. Job titles in this space often read 'junior computational scientist', 'scientific software developer' or 'modelling engineer'.
Specialist MSc as a bridge — courses in computational science, scientific computing, HPC, computational chemistry, bioinformatics, computational biology or data science are a recognised way in if your first degree was lab-based or purely theoretical.
Graduate schemes at large engineering, energy, defence and aerospace employers, where you rotate through simulation and analysis teams. These are less common than general engineering or data schemes but they do exist.

Show 2 more

Internships, summer placements and research assistantships — a summer in a university research group, a placement year with an engineering or pharma modelling team, or a student placement at a research facility. Demonstrable project work carries a lot of weight here.
Sideways moves from software engineering or data science into a scientific domain, or from experimental science into modelling. Both happen regularly; the harder gap to close is usually the one you didn't train in.

What employers ask for

A numerate degree is essentially non-negotiable — physics, chemistry, maths, engineering, computer science, biochemistry, earth sciences or similar. The exact subject matters less than whether you can show the relevant maths and the relevant coding.
Grades: a 2:1 or above is the usual expectation, with many research-oriented employers in practice recruiting at first-class/distinction level or via a PhD. Some commercial software roles care more about your code than your classification.
A PhD is frequently required or strongly preferred for roles involving developing new methods, and much less so for roles applying existing tools. Read the advert carefully — the two kinds of job share a title.
Evidence of real programming, not just a module you passed: a GitHub repository, a dissertation codebase, contributions to an open-source scientific package, or a placement where you shipped working code.

Show 2 more

Domain knowledge in the employer's science — e.g. quantum chemistry and molecular dynamics for drug discovery, fluid dynamics for aerospace, genomics and statistics for bioinformatics. Employers vary on whether they expect this up front or will train you.
Security clearance is a requirement for some defence, nuclear and government laboratory roles, which usually means UK residency history and can lengthen the start date considerably. There are no formal chartered qualifications for the role, though engineering-sector computational scientists sometimes work towards Chartered Engineer status.

Skills that matter

Scientific programming in Python plus at least one compiled language

Python handles analysis, glue code and plotting, while C++, Fortran or similar is where the performance-critical numerical work lives — most teams expect you to be comfortable moving between them.

Numerical methods and applied maths

You need to know why a solver diverged, what a timestep or grid resolution is doing to your answer, and when a result is numerical noise rather than science.

Working on Linux and HPC clusters

Nearly all large-scale simulation runs on Linux via a command line and a job scheduler, so comfort with shell scripting and remote working is a daily practical necessity.

Statistical and data handling skills

Simulations produce large, messy outputs, and distinguishing a genuine effect from sampling error or a configuration mistake is a core part of the job.

Software engineering discipline — version control, testing, documentation

Scientific code is often used for years by people who didn't write it, and results that can't be reproduced are worth very little to a regulator, a reviewer or a product team.

Translating between scientists and computation

Much of your value comes from turning 'can we tell whether this molecule will be stable?' into a well-posed model, and then explaining the uncertainty in the answer to someone who will make a decision on it.

Putting these on a CV? We counted which skills 12,914 live graduate adverts ask for, and what to do with the answer — part of our application advice.

Where it leads

Junior/graduate computational scientist: running established models and pipelines, supporting senior colleagues' projects, learning the domain science and the codebase.
Computational scientist: owning projects end to end — choosing the method, building the model, defending the results to the people who commissioned them.
Senior computational scientist: setting technical direction for a problem area, mentoring juniors, deciding which approaches the team invests in. Timescales here vary a lot — a few years in fast-moving commercial teams, longer in research institutions where progression is tied to publications or funding.
A fork in the road: a technical track (principal or lead scientist, method or software architecture ownership) or a management track (team lead, group leader, head of modelling).
Common sideways destinations: data science and machine learning engineering, quantitative finance, scientific software product roles, consultancy, academic research fellowships and lectureships, or specialised technical roles at national research facilities.

What people get wrong

“It's basically data science with a lab coat on.”

Data science mostly infers patterns from observed data; computational science mostly generates data from a model of the underlying physics, chemistry or biology. The maths, the tools and the failure modes are different, although machine learning is now genuinely woven into both.

“You spend your days thinking deep thoughts about equations.”

A large share of the week is unglamorous engineering — fixing builds, chasing down why a cluster job died at hour nine, cleaning up data formats and making someone's decade-old code compile again. The science is real, but it arrives in small windows between plumbing.

“If you don't have a PhD, the door is shut.”

A PhD is the norm for method-development and research-lab roles, but there is a substantial and growing set of jobs — scientific software development, simulation support, modelling in industry — that hire strong graduates and MSc holders on the strength of their coding and numerical ability.

“The job is all about getting the biggest, fastest simulation possible.”

In industry the usual constraint is a decision deadline, not compute power. Knowing when a crude, fast approximation is good enough — and being able to say honestly how wrong it might be — is valued more than maximum fidelity.

Where this varies

The title covers quite different jobs. In pharma and biotech it usually means molecular modelling, cheminformatics or bioinformatics supporting drug discovery, and PhDs dominate. In aerospace, automotive, nuclear and energy it usually means CFD (computational fluid dynamics) or structural/thermal simulation, often sitting inside an engineering team and sometimes on a path to chartered engineer status. At national labs, research facilities and universities it leans towards method development, open-source code and publication. At scientific software companies and HPC centres it is closer to software engineering with a science background, and your code quality matters more than your publication record. Geographically, the work clusters around Cambridge, Oxford, London, the Harwell and Daresbury research campuses, Edinburgh, Manchester, Bristol and the aerospace and nuclear corridors in the North West and South West — remote working is more available in the software-leaning roles than in cleared defence or facility-based ones.

See all 2 live Computational Scientist jobsFrom employers' own careers sites, updated daily Tailor your CV for this roleFree with an account, 3 a month

General guidance about the role across the UK market, not about any specific employer. Entry routes and requirements vary — always check the individual job advert.