Most data scientist CVs describe models instead of outcomes. "Built a random forest classifier" appears in hundreds of data scientist applications — and it tells a hiring manager nothing about whether the model was accurate, deployed, or useful. The CVs that get shortlisted describe what problem the model solved, what data it learned from, what the evaluation metrics were, and what the business actually changed as a result of using it. That shift — from describing a technical exercise to demonstrating a business impact — is the single biggest change a data scientist can make to their CV. The second most common failure is the absence of SQL, which signals to any experienced hiring manager that the candidate works primarily on pre-cleaned datasets and has not encountered the messy extraction work that constitutes 60% of most data science roles. This guide covers both, along with the ATS keywords, CV structure, and specific mistakes that filter data scientist applications before any human reads them.

What Data Scientist Job Descriptions Actually Require in 2026

Data scientist job descriptions in 2026 have diverged into two archetypes, and misidentifying which one you are applying for is one of the most common job search mistakes in the discipline.

Product/business data scientist — spends the majority of time on SQL queries, experimental design, and statistical analysis; builds machine learning models for specific business problems (churn prediction, recommendation systems, fraud detection); presents findings to non-technical stakeholders. This is 70-80% of "data scientist" job postings.

ML engineer / applied scientist — builds and deploys ML models at scale; writes production Python or Java code; owns the model training pipeline and serving infrastructure; collaborates with SWE teams on integration. This is 20-30% of postings, usually at larger tech companies.

Requirements across both archetypes:

  • Python — pandas, NumPy, scikit-learn for classical ML; PyTorch or TensorFlow for deep learning; Matplotlib/Seaborn/Plotly for visualisation. All expected at mid-level.
  • SQL — the most commonly overlooked skill on data scientist CVs. Every data science role requires extracting data from databases before analysing it. "SQL for complex data extraction, window functions, and performance optimisation" belongs in your Skills section.
  • Statistical knowledge — hypothesis testing, confidence intervals, A/B test design, regression analysis, probability distributions. Data science interviews probe statistics harder than most candidates expect.
  • Machine learning fundamentals — feature engineering, cross-validation, hyperparameter tuning, model evaluation metrics (not just accuracy — precision, recall, F1, AUC-ROC, RMSE depending on the problem).
  • Communication — presenting model results and uncertainty to non-technical stakeholders; JDs phrase this as "ability to translate data insights into business recommendations."

Data scientist salaries in 2026: £50K–£90K UK; $110K–$175K US for product data scientists. Applied scientists and ML engineers at FAANG-adjacent companies exceed $200K total compensation.

ATS Keywords for a Data Scientist CV

Data scientist ATS searches are keyword-specific at both the tool level and the method level. Hiring managers filter by tool (Python, SQL, PyTorch) AND by methodology (A/B testing, feature engineering, time series forecasting). Both layers need to appear in your CV.

Essential ATS terms for a data scientist CV:

  • Title variants: Data Scientist, Machine Learning Engineer, ML Engineer, Applied Scientist, Research Scientist, NLP Engineer, Computer Vision Engineer
  • Core languages: Python, R, SQL, Scala
  • ML and frameworks: scikit-learn, PyTorch, TensorFlow, Keras, XGBoost, LightGBM, Hugging Face, Spark, PySpark
  • Data tools: pandas, NumPy, Jupyter, Databricks, BigQuery, Redshift, dbt, Airflow, MLflow
  • Methods: machine learning, deep learning, natural language processing, computer vision, A/B testing, statistical modelling, feature engineering, time series forecasting, recommendation systems, anomaly detection
  • Cloud and MLOps: AWS SageMaker, GCP Vertex AI, Azure ML, MLflow, Weights & Biases, Docker
  • Long-tail phrases: data scientist resume examples, how to write a data scientist resume, data scientist cv template, data scientist skills for resume

Placement: Python and SQL belong in your headline alongside your title. List methods (A/B testing, NLP, etc.) in Skills as a separate "Methods" category — this is the keyword layer most data scientist CVs omit. ATS systems searching for "A/B testing" will only match your CV if the phrase appears explicitly.

Data Scientist CV Structure and Bullets That Show Business Impact

Section order:

  1. Headline — "Data Scientist | Python · SQL · PyTorch · A/B Testing" or "Data Scientist | Machine Learning · NLP · Python · BigQuery"
  2. Skills — layered: Languages → ML Frameworks → Data Tools → Methods → Cloud/MLOps
  3. Experience — 4–6 bullets per role; model outcomes and business impact here
  4. Education — higher placement than in other engineering roles; MSc or PhD in Statistics, Mathematics, or CS is valuable and often required for research-adjacent roles
  5. Publications or Projects — if you have published research (even preprints on arXiv), list it; for non-research roles, 2–3 Kaggle or personal projects with notable outcomes

Two pages for 4+ years; one to two pages for under 4.

Strong data scientist bullet points need: the model or method used, the data it operated on (scale matters), the evaluation metric achieved, and the business outcome or deployment status. Three examples:

  • Built and deployed a customer lifetime value prediction model using XGBoost on 3 years of transactional data (18M rows), achieving 76% accuracy within ±20% of actual CLV; model integrated into the CRM and improved marketing spend allocation, driving a 14% improvement in campaign ROI
  • Designed and analysed 6 A/B experiments across the checkout funnel — including power analysis, sample size calculation, and Bayesian significance testing — with findings directly influencing 3 product decisions projected to deliver £2M+ annual impact
  • Developed a real-time fraud detection pipeline in Python with PySpark on AWS EMR, classifying 400K daily transactions with sub-200ms latency; reduced false positive rate from 3.2% to 0.8% over the previous rule-based system, cutting manual review workload by 65%

Data scientist interviews typically involve: a take-home modelling challenge (given a dataset, build and evaluate a model), a live SQL test (2–4 queries of increasing complexity), statistics questions (explain p-values, walk me through A/B test design), and a business case discussion (how would you approach this problem). Your CV determines the difficulty and domain of the modelling challenge — production deployment evidence leads to harder, more interesting questions.

Three Data Scientist CV Mistakes That Cost You Interviews

Model descriptions without outcomes. This is the defining weakness of most data scientist CVs. "Developed a neural network for product recommendation" is technically descriptive but professionally meaningless. A hiring manager cannot tell if the model was 60% accurate or 96% accurate, whether it was ever deployed, or whether anyone used it. The fix is always the same: add the evaluation metric, the deployment status, and one business metric. "Developed a collaborative filtering recommendation model in PyTorch — deployed to production and responsible for 23% of total user engagement within 90 days of launch" is a complete bullet.

SQL absent from the CV. The ability to write SQL is assumed to be universal among data scientists — but it is also filtered for explicitly in ATS systems and asked about directly in phone screens. A CV that lists Python, PyTorch, and NumPy but not SQL looks incomplete to experienced hiring managers who know that 60% of data science work is data extraction and preparation. SQL belongs in your Skills section as a primary skill, not buried or omitted.

Kaggle competitions treated as primary experience. A top-10 Kaggle finish demonstrates modelling on clean, labelled, competition-optimised data — not the messy, ambiguous, stakeholder-dependent data of business data science. Recruiters know this distinction. List Kaggle in a Projects section as supporting evidence of technical skill; do not lead with it or position it as equivalent to production work. One strong production outcome bullet outweighs three Kaggle medals for most hiring purposes.

If you are applying to data scientist roles and want your CV rebuilt around the specific Python, SQL, and machine learning requirements in a target job description, Resumegpt generates your data scientist CV from your work history in under 60 seconds — ATS-optimised, model outcomes included, and exported as a PDF ready to submit.