Listen to this article
If you are writing a data science CV for a UK or European employer, put your projects where a recruiter will see them in the first half page. Treat that section as a compact portfolio with results in it. Give published papers, Kaggle placings or your thesis their own clearly labelled space, and group your tools by what you use them for instead of listing thirty names in alphabetical order. The tech stack matters less than most candidates think. What a reader actually looks for is evidence that you framed a problem, built something and measured whether it worked.
Why a data science CV is not an IT resume
We already have a piece on what recruiters scan for on an IT resume, and much of it holds for engineers who ship features. Data science work has a different shape. An engineer's output is usually a system, and it either exists or it doesn't. Yours is often a question that got answered, a model that beat a baseline, or an analysis that changed a decision.
To judge that kind of work, the reader needs three things: context (what the problem was and what data you had), method (what you tried and why) and outcome (what changed). A list of jobs with duties under them can't hold any of that. So a data science CV has more sections that matter: experience, a projects section that works like a mini-portfolio, and for many candidates a publications or competitions section as well.
Data scientist resume format: the order that works in the UK and EU
In the UK and most of Europe, the CV is the whole document. It usually runs to two pages and, for UK roles, has no photo (our guide to the UK CV format covers the other conventions). This order suits most data roles:
- Name, city, email, phone, and one or two links: GitHub, a portfolio site, Google Scholar or your Kaggle profile.
- A three-line profile: what kind of data scientist you are, which domain you work in, and one proof point.
- Experience, with project-level detail inside each role.
- Selected projects. Move this above Experience if you are a graduate or coming from an analyst role.
- Publications, competitions or thesis.
- Skills, grouped.
- Education, with your dissertation title if it is relevant.
The rule for the order is simple: your strongest evidence goes highest. For a new MSc graduate with a solid dissertation and a Kaggle medal, that means projects and publications, not a three-month retail job.
How to list machine learning projects on a CV

Each project gets a title, one line of context and two or three bullets. A pattern that works is problem, data, approach, result and what happened next.
A typical weak version reads: "Built a machine learning model using Python, scikit-learn and XGBoost to predict customer churn." It names tools, but it tells the reader nothing about your judgement.
Here is the same project rewritten, as an illustration:
Customer churn prediction, anonymised telecom data (MSc capstone)
- Framed churn as a 90-day classification problem and handled heavy class imbalance with stratified sampling and cost-sensitive evaluation instead of raw accuracy.
- Compared logistic regression with gradient boosting and chose the simpler model, because it came close on recall and the industry partner could understand its drivers.
- Presented the findings to the partner's retention team, who used the ranked customer list in a pilot.
Use your own numbers and outcomes, and only ones that are true. The value comes from the structure: each bullet shows a decision.
What counts as a result without a business metric
Academic and personal projects rarely come with revenue figures, and you don't need them. A result can be an improvement over a stated baseline, a pipeline someone else can rerun, a finding that went against the obvious hypothesis, or a problem you caught. "Identified and removed a leaking feature that had inflated validation scores" tells a hiring manager more than most accuracy figures do.
Where your data science portfolio goes on the CV
Link each project to its repository on the title line, and write the address out as plain text, because some application systems strip hyperlinks. Make sure the README explains the problem and the result before anyone opens a notebook. If you are not sure how to format links safely, see how to put a portfolio link on your CV without breaking ATS. A recruiter may never click, but a technical interviewer often will, so the repo needs to back up what the bullet claims.
Do not bury the paper or the production model
This is the most costly mistake we see on data CVs. Someone has a model making daily predictions that a real team relies on, and their CV says "Responsible for data analysis and reporting across the business." A recruiter reads that as a reporting role and moves on. The most valuable thing the candidate has done has disappeared into a phrase that fits thousands of other CVs.
Name what you built, how it runs and who uses it: "Built and deployed a demand forecasting model, retrained weekly through a scheduled pipeline, which the planning team uses to set stock orders." Now the reader can see a production system, a deployment skill and a stakeholder in one line.
The same thing happens with research. A thesis accepted at a workshop, or a co-authored journal paper, often ends up as half a line under Education. Give it its own heading and mention it in your profile too.
Publications, Kaggle and the thesis
- Publications: use a standard citation format, put your own name in bold, and include the venue and year. Label preprints as preprints.
- Kaggle: include medals, competitive placings with an honest rank or percentile, or notebooks other people genuinely used. Say whether you worked solo or in a team. Introductory competitions are not credentials.
- Thesis or capstone: give the title, one line on what you found and one on the method. Add the grade if it helps you.
Why the tech stack section is overrated
An alphabetical list ("Airflow, AWS, BigQuery, Docker, Excel, Git, Hadoop, Keras...") says nothing about depth. Most candidates' lists look almost identical, so the section rarely sets anyone apart. A list that includes every library you have imported once also sets you up for interview questions you can't answer.
The tools that carry weight are the ones that appear in your bullets, attached to real work. Treat the skills section as a reference and a keyword net for screening software, not as your argument. Grouping it by role shows a reader the shape of your skills in seconds:
- Modelling: Python (scikit-learn, PyTorch), statistical testing, time series forecasting
- Data and pipelines: SQL (PostgreSQL, BigQuery), dbt, Airflow
- Deployment: Docker, MLflow, a cloud ML platform you have actually used
- Visualisation and communication: Tableau, matplotlib, written reporting for non-technical teams
Only list what you would happily discuss for ten minutes. Where the advert's wording is accurate for you, use it.
Data analyst vs data scientist CV: what changes
An analyst CV leads with SQL, dashboards and the decisions your analysis supported. A data scientist CV also has to show modelling, experimentation and how you handled uncertainty. If you are moving across, keep your analyst experience but rewrite it around the method you used. "Produced weekly sales dashboards" becomes "Designed the analysis for a pricing A/B test and built the dashboard the commercial team used to track it."
Then add one or two modelling projects to cover the gap. Keep your real job title. Let the profile say where you are heading, and let the projects show you can already do the work.
What live UK data science postings suggest for your CV
According to Job Opportunities API data as of 11 November 2026, there are 846 live data science openings in the UK, and 123 of them (14.5%) were posted in the last seven days. London has 443 of them, far ahead of Manchester with 23 and Bristol with 21. The employers with the most live openings are Wise (20), RELX (18), Faculty (17), JPMorgan Chase (16) and PA Consulting (15).
That list mixes product and financial companies with consultancies, and it affects how you order your projects. A consultancy reader will probably look for range and for proof that you can explain your work to a client. A product company reader will probably look for something that runs in production. You can use the same projects for both, in a different order and with a different first bullet. That is tailoring, not a rewrite. With roughly one posting in seven new this week, it is worth keeping two versions ready instead of rebuilding from scratch each time.
Only 77 postings (9.1%) carry a remote status stated by the employer, so put a real location on your CV and don't build it around remote work. JOA detected a pay figure in at least 10.6% of these postings. That is a lower bound, because detection undercounts, so the real share is higher.
Germany shows 1,004 live openings, with 260 (25.9%) posted in the last seven days and Berlin leading with 100. Accenture alone accounts for 368, so consultancy-style framing matters even more there. These figures are today's openings, not hires and not a forecast.
Data: Job Opportunities API (jobopportunitiesapi.org), live job postings as of 11 November 2026.
Before you send it
If you have sent a dozen applications and heard nothing back, it's easy to conclude the market is against you. Sometimes the cause is simpler: your best work is on page two, under a heading nobody reads. A real person will read your CV for free and tell you what works, what raises doubts and what to fix first, and if it's already strong, they will say so. Get my free CV review


