Data scientist jobs: skills, pay and how to get hired
Updated 5 Nov 2026. Figures from the jobs in our search, not the whole market.
Contents
See open Data Scientist jobsScore your CV for these jobs
What a data scientist does
A data scientist turns a business question into an answer the business can act on: which customers are likely to leave, how many parts to stock next month, whether a new price works better than the old one. The work is finding and cleaning the data, choosing a method, building and testing a model, and explaining the result to people who will make a decision with it.
Job titles vary: Data Scientist, Applied Scientist, Machine Learning Scientist, and senior titles such as Lead or Principal Data Scientist. Some roles lean towards statistics and experiments, others towards building machine learning models that run in a product. We count the general titles as one role.
The skills the ads ask for
We read a sample of the ads every day and count the skills they name. On a computer, hover a skill in the text to see its share; on a phone, the table has the same figures.
| Skill | Share of ads | What is expected |
|---|---|---|
| Python | 62% | Python for data: clean, tested notebooks and scripts, with pandas, scikit-learn and plots others can rerun. |
| SQL | 42% | Pull and shape your own data: joins, aggregations and window functions, on large tables. |
| Machine Learning | 52% | Train, validate and compare models: splits, cross-validation, metrics, overfitting and leakage. |
| Statistics | 20% | Distributions, confidence intervals, hypothesis tests and their traps; enough to tell a real effect from noise. |
| Deep Learning | 8% | Neural networks for images, text or signals; know when they beat simpler models and when not. |
| Pandas | 5% | Load, clean, join and reshape tables in Python; know where it gets slow on big data. |
| Spark | 7% | Process data too big for one machine, often on Databricks or a cloud platform. depends on the provider |
| PyTorch | 15% | The most common deep learning framework in new work: build, train and fine-tune neural networks. depends on the release |
| TensorFlow | 12% | One of the two main deep learning frameworks; common in older production systems. depends on the release |
| MLOps | 11% | Ship and run models: versioned data and models, automated retraining, monitoring and rollbacks. |
| Docker | 7% | Package a model service with its dependencies so it runs the same everywhere. |
| AWS | 17% | Storage, compute and SageMaker for training and serving models. depends on the provider |
| Azure | 16% | Microsoft's cloud, with Azure Machine Learning; common in larger European firms. depends on the provider |
| Google Cloud | 7% | Google's cloud, with BigQuery and Vertex AI for data and models. depends on the provider |
| Airflow | 7% | Schedule and monitor data pipelines, such as a nightly retraining job. |
| Data Visualization | 7% | Charts that answer one question each, read correctly by people who are not analysts. |
| Tableau | 2% | A dashboard tool many business teams use to read your results. depends on the provider |
| Power BI | 12% | Microsoft's dashboard tool, common where the company runs on Microsoft. depends on the provider |
| NLP | 4% | Natural language processing: classify, extract from and search text, now often with language models. |
| LLM | 25% | Use large language models through an API for text tasks, and measure them like any other model. |
| Git | 16% | Version control for code and notebooks; reviewed changes, not files called final_v3. |
| English | 60% | The working language of many data teams, and of the papers and documentation you will read. |
Shares of the data scientist ads we read in the last 7 days, as of 5 Nov 2026. Our ads, not the whole market.
Three things stand out. First, PythonAsked in 62% of data scientist ads (as of 5 Nov 2026).Python for data: clean, tested notebooks and scripts, with pandas, scikit-learn and plots others can rerun. and SQLAsked in 42% of data scientist ads (as of 5 Nov 2026).Pull and shape your own data: joins, aggregations and window functions, on large tables. are the everyday tools: most of the job is getting and shaping data. Second, machine learningAsked in 52% of data scientist ads (as of 5 Nov 2026).Train, validate and compare models: splits, cross-validation, metrics, overfitting and leakage. and statisticsAsked in 20% of data scientist ads (as of 5 Nov 2026).Distributions, confidence intervals, hypothesis tests and their traps; enough to tell a real effect from noise. go together; ads want both, because a model is only useful if you can tell whether it really works. Third, the tools for bigger data and for running models, such as SparkAsked in 7% of data scientist ads (as of 5 Nov 2026).Process data too big for one machine, often on Databricks or a cloud platform. depends on the provider, MLOpsAsked in 11% of data scientist ads (as of 5 Nov 2026).Ship and run models: versioned data and models, automated retraining, monitoring and rollbacks. (shipping and running models) and a cloud such as AWSAsked in 17% of data scientist ads (as of 5 Nov 2026).Storage, compute and SageMaker for training and serving models. depends on the provider or AzureAsked in 16% of data scientist ads (as of 5 Nov 2026).Microsoft's cloud, with Azure Machine Learning; common in larger European firms. depends on the provider, appear in fewer ads but often decide senior interviews.
Deep learning frameworks such as PyTorchAsked in 15% of data scientist ads (as of 5 Nov 2026).The most common deep learning framework in new work: build, train and fine-tune neural networks. depends on the release matter for roles with images, text or speech. Many roles also expect you to show results in a dashboard tool such as TableauAsked in 2% of data scientist ads (as of 5 Nov 2026).A dashboard tool many business teams use to read your results. depends on the provider or Power BIAsked in 12% of data scientist ads (as of 5 Nov 2026).Microsoft's dashboard tool, common where the company runs on Microsoft. depends on the provider.
Where the jobs are, country by country
Each country below has its own figures: the cities with the most jobs in our list when a few clearly lead, whether the number is growing, how often ads offer remote or hybrid work, and pay ranges from published research, in the local currency.
Germany
18% of the ads offer hybrid work, and 7% are fully remote.
Pay (Base salary)
- Entry level: €48,000 to €63,000 a year
- Mid-level: €60,000 to €80,000 a year
- Senior: €78,000 to €105,000 a year
Munich and Stuttgart pay the most among cities; Baden-Württemberg, Hesse and Hamburg by state.
Sources: gehalt.de (2026), Indeed (Oct 2026), Payscale (Sep 2026).
Pay researched on 27 Oct 2026.
Language: many ads in Germany mention GermanAsked in 62% of data scientist ads (as of 5 Nov 2026).Named in many ads in Germany, Austria and Switzerland, often for presenting results to the business., often because results are presented to business teams. International teams work in EnglishAsked in 60% of data scientist ads (as of 5 Nov 2026).The working language of many data teams, and of the papers and documentation you will read.; the ad usually says which.
Law (EU): GDPRAsked in 6% of data scientist ads (as of 5 Nov 2026).Personal data in training sets: purpose, minimisation, pseudonymised data and the right to deletion. applies to personal data in training sets, and the EU AI Act adds rules for models used in hiring, credit and other high-risk uses.
Job figures as of 5 Nov 2026.
United States
20% of the ads offer hybrid work, and 48% are fully remote.
100% of the ads that say are full-time.
Pay (Base salary)
- Entry level: $85,000 to $115,000 a year
- Mid-level: $115,000 to $155,000 a year
- Senior: $150,000 to $200,000 a year
San Francisco and Seattle pay the most; big tech companies add stock on top of these ranges.
Sources: US Bureau of Labor Statistics (May 2025 data), Built In (2026), Indeed (Oct 2026), Robert Half 2027 Salary Guide (a projection).
Pay researched on 27 Oct 2026.
Job figures as of 5 Nov 2026.
United Kingdom
Most data scientist jobs in our list are in London, Cardiff and Manchester.
29% of the ads offer hybrid work, and 43% are fully remote.
Pay (Base salary)
- Entry level: £32,000 to £50,000 a year
- Mid-level: £50,000 to £80,000 a year
- Senior: £75,000 to £100,000 a year
London pays somewhat more than the rest of the UK.
Advertised salaries run well above what people report in surveys, so read both ends.
Sources: ITJobsWatch (Oct 2026), ITJobsWatch (senior) (Oct 2026), Indeed (Oct 2026), Payscale (Jul 2026), Robert Half 2027 Salary Guide (a projection).
Pay researched on 27 Oct 2026.
Job figures as of 5 Nov 2026.
Switzerland
Pay (Base salary)
- Entry level: CHF 85,000 to CHF 105,000 a year
- Mid-level: CHF 100,000 to CHF 130,000 a year
- Senior: CHF 125,000 to CHF 155,000 a year
Zurich pays the most. The Swiss sources rest on small samples, so read them as a rough guide.
Sources: Robert Half 2027 Salary Guide (a projection), Indeed (Oct 2026), Payscale (Oct 2025).
Pay researched on 27 Oct 2026.
Language: ads around Zurich and Basel often ask for German, ads around Geneva and Lausanne for French; many data teams work in English.
India
Pay (Base salary)
- Entry level: ₹5 lakh to ₹8 lakh a year
- Mid-level: ₹8 lakh to ₹15 lakh a year
- Senior: ₹14 lakh to ₹25 lakh a year
Hyderabad, Mumbai and Gurugram pay the most; multinationals and product companies pay above these ranges.
Sources: Indeed (Oct 2026), Indeed (senior) (Sep 2026), Payscale (Sep 2026).
Pay researched on 27 Oct 2026.
Your CV for a data scientist job
A data scientist CV is read twice: first by the hiring software, which looks for the words of the ad, then by a data scientist or manager, who looks for business results. Write for both.
Use the words the ads use
- Name methods and tools as the ad does: "machine learning", "A/B testing", "Python", "SQL", "PyTorch". Add the long form once if it helps.
- Put a short skills line near the top: PythonAsked in 62% of data scientist ads (as of 5 Nov 2026).Python for data: clean, tested notebooks and scripts, with pandas, scikit-learn and plots others can rerun., SQLAsked in 42% of data scientist ads (as of 5 Nov 2026).Pull and shape your own data: joins, aggregations and window functions, on large tables., machine learningAsked in 52% of data scientist ads (as of 5 Nov 2026).Train, validate and compare models: splits, cross-validation, metrics, overfitting and leakage., statisticsAsked in 20% of data scientist ads (as of 5 Nov 2026).Distributions, confidence intervals, hypothesis tests and their traps; enough to tell a real effect from noise., your cloud and your dashboard tool.
- Say what kind of data scientist you are: experiments and statistics, models in production, or language and images. Ads differ a lot.
- Only list what you could discuss for ten minutes. Every method on the CV can become an interview question.
Prove each skill
A method in a list proves nothing. A line under a job proves it: the question, what you did, and the business result, such as money saved, customers kept or hours saved.
| Skill | Weak line | Strong line, with proof |
|---|---|---|
| Machine learning | Built machine learning models. | Built a churn model for 400,000 customers; the retention team's calls to the top 5% saved about €1.2m a year. |
| Statistics | Strong statistics. | Designed and analysed 14 A/B tests of the checkout; the 3 winners lifted conversion by 6%. |
| SQL | SQL. | Rebuilt the forecasting data set in SQL from 6 sources; forecast error fell from 18% to 11%. |
| MLOpsAsked in 11% of data scientist ads (as of 5 Nov 2026).Ship and run models: versioned data and models, automated retraining, monitoring and rollbacks. | Deployed models. | Put the demand model into production with weekly retraining and drift alerts; no silent failures in a year. |
| Python | Python, pandas, scikit-learn. | Replaced a manual Excel pricing process with a Python pipeline; a day of work a week saved. |
Never claim what you did not do. A side project or a thesis counts: say what it is, link the code, and describe its results the same way.
How we read our data
- The jobs: the data scientist jobs in our job search, matched the same way the search matches them. Our list is not the whole market, so we give shares and rankings, never counts.
- The skills: we read a sample of the ads every day and count the skills in them with a fixed word list. We keep only the counts, never the ad text.
- Pay: from published salary research, with each source and its date, as a range from entry level to senior. It is not worked out from our ads.
- Dates: every figure is worked out again each day, and shows the day it was worked out.
Frequently asked questions
- What does a data scientist do?
- Uses data, statistics and machine learning to answer business questions and build predictions, then explains the result so people can act on it.
- What is the difference between a data scientist and a data analyst?
- Both work with data. Analysts mostly describe what happened and why, with SQL and dashboards; data scientists more often predict what will happen, with statistics and machine learning.
- Do I need a PhD to become a data scientist?
- No, most ads do not ask for one. A degree in a numerate subject is common; what counts most is proof that you can solve a real problem with data.
- What should a data scientist CV show first?
- Two or three projects with the business question, the method and a measured result, then a short skills line in the words of the ad.
More on the data scientist role
The cvscored team
See open Data Scientist jobsScore your CV for these jobs
Related roles
Jobs from Arbeitnow and the Bundesagentur für Arbeit