The credibility gap in data science CVs is between models built and models used. Almost every applicant can describe a model; far fewer can describe one that runs on live data, has a monitored performance metric, and changed how a business operates. The second kind of experience is what most job adverts are actually trying to buy.
Evaluation honesty is the other signal reviewers weigh heavily. A CV that quotes only accuracy on an imbalanced problem tells an experienced interviewer more than the applicant intended. Naming the metric that suits the problem - precision at a fixed recall, AUC-PR, calibration - demonstrates the judgement the role requires.
This example is written for a data scientist with five years of experience who has taken models into production in a commercial setting.
Opens in the editor with this content already filled in, so you can replace it with your own.
Rohan Kapoor
Data Scientist · Predictive Modelling & Experimentation
rohan.kapoor@example.com
+91 98110 55234
Delhi, India
linkedin.com/in/example-rohan-kapoor
github.com/example-rohankapoor
Profile
Data scientist with five years in retail and financial services. Have taken three models into production, including a churn model feeding the retention team’s weekly call list and a demand elasticity model validated with a switchback test showing a 2.4% margin improvement. MSc in Statistics.
Professional Experience
Data ScientistApril 2022 – Present
Ambleside Retail Analytics · Delhi, India
•Built a gradient-boosted churn model achieving 0.71 precision at 60% recall against a 0.34 baseline, deployed as a weekly batch score feeding the retention team’s call list.
•Developed a demand elasticity model across 1,800 products and validated it with a four-week switchback test, which showed a 2.4% margin improvement with no measurable volume loss.
•Set up drift monitoring on the six most influential features and a monthly recalibration job after the churn model degraded from 0.71 to 0.58 precision within two quarters.
•Rewrote the feature pipeline in Spark, cutting the training data build from 6 hours to 40 minutes and making weekly retraining practical.
•Present model readouts to the commercial director monthly, including the cases where the recommendation is to do nothing.
Python · XGBoost · Spark · MLflow · SQL
Junior Data ScientistAugust 2020 – March 2022
Sentinel Credit Services · Gurugram, India
•Built the early-arrears propensity model used to prioritise collections contact, improving contact-to-recovery conversion by around 15%.
•Produced the model documentation pack required by the internal model risk committee, including limitations, monitoring plan and fairness analysis across age and region.
•Automated the monthly performance report for four production models, replacing a manual notebook run.
Python · scikit-learn · SQL · Airflow
Selected Projects
Retail footfall nowcasting
Lead
•Combined till transactions, weather and local event data to estimate same-day footfall for 60 stores.
•Reduced staffing over-allocation on quiet days by around 9% in the pilot region.
Education
MSc StatisticsJuly 2018 – June 2020
University of Delhi · Delhi, India
•Thesis on survival models for subscription retention.
Distinction
BSc (Hons) MathematicsJuly 2015 – May 2018
University of Delhi · Delhi, India
First Class
Publications
Practical calibration for imbalanced retail churn modelsNovember 2023
Kapoor, R. and Menon, K. · Indian Conference on Applied Analytics (Poster)
Data Scientist example on the Data Professional layout. All details are fictional and shown for demonstration only.
What recruiters expect
Before writing anything, it helps to know what the person reading is checking for. In this field that is usually a short, specific list:
At least one model in production, with the serving pattern and the monitoring described.
A metric appropriate to the problem, and awareness of the baseline it beat.
Experiment design: how the model was validated in the real world, not only in cross-validation.
Solid SQL and data engineering awareness, because most of the work is upstream of the model.
Communication of uncertainty to non-technical decision makers.
Recommended CV structure
This is the running order the example uses. It is a starting point rather than a rule, but the order reflects what tends to be read first in this profession.
Profile — Three or four lines positioning you for the role.
Technical Skills — Grouped skills, for example "Languages" and "Tooling".
Professional Experience — Paid roles, in reverse chronological order.
Selected Projects — Work you built, with outcomes and the stack used.
Education — Degrees, diplomas and school-leaving qualifications.
Publications — Papers, articles and posters, in your citation style.
Certifications — Completed certifications with the issuing body.
Languages — Spoken languages with CEFR levels.
Sections worth adding
Talks — Conference presentations substitute well for publications in industry roles.
Teaching — Relevant if you have supervised students or run internal training.
Skills worth including
Grouped rather than listed in one block. Grouping makes a long list readable and shows that you can tell the difference between the things you use daily and the things you have touched.
Jupyter · Quarto · Plotly · Executive readouts · Model cards
Beyond the technical list: Framing an ambiguous business question, Presenting uncertainty honestly, Working with data engineers, Peer review of analysis. These belong inside your experience bullets, demonstrated, rather than in a list of adjectives.
Example professional summary
Three or four lines, positioned for the role rather than describing your personality. Two versions you can adapt:
Data scientist with five years in retail and financial services. Have taken three models into production, including a churn model feeding the retention team’s weekly call list and a demand elasticity model validated with a switchback test that showed a 2.4% margin improvement. MSc in Statistics.
Applied data scientist focused on the parts of the job that survive contact with production: feature pipelines, honest evaluation and monitoring. Comfortable presenting an uncertain answer to a commercial audience.
Writing your experience
The difference between a CV that gets a call and one that does not is almost always in the bullet points. Each pair below shows a real rewrite of the kind of line that appears on most CVs in this field.
Weak
Built a churn prediction model with 92% accuracy.
Stronger
Built a gradient-boosted churn model achieving 0.71 precision at 60% recall against a 0.34 baseline, deployed as a weekly batch score feeding the retention team’s call list.
Uses metrics that mean something for an imbalanced problem, names the baseline, and shows where the output goes.
Weak
Used machine learning to improve pricing.
Stronger
Developed a demand elasticity model across 1,800 products and validated it with a four-week switchback test, which showed a 2.4% margin improvement with no measurable volume loss.
Real-world validation design, not just offline evaluation.
Weak
Monitored model performance.
Stronger
Set up drift monitoring on the six most influential features and a monthly recalibration job after the original model degraded from 0.71 to 0.58 precision within two quarters.
Admits a model degraded, which is normal and signals genuine production experience.
Taken from the example
The sample CV for this profession is fully written. A few sections from it, so you can see the level of specificity that works:
Experience
Data Scientist, Ambleside Retail Analytics
Built a gradient-boosted churn model achieving 0.71 precision at 60% recall against a 0.34 baseline, deployed as a weekly batch score feeding the retention team’s call list.
Developed a demand elasticity model across 1,800 products and validated it with a four-week switchback test, which showed a 2.4% margin improvement with no measurable volume loss.
Set up drift monitoring on the six most influential features and a monthly recalibration job after the churn model degraded from 0.71 to 0.58 precision within two quarters.
Rewrote the feature pipeline in Spark, cutting the training data build from 6 hours to 40 minutes and making weekly retraining practical.
Projects
Retail footfall nowcasting — Combined till transactions, weather and local event data to estimate same-day footfall for 60 stores.
Education
MSc Statistics, University of Delhi — Distinction
BSc (Hons) Mathematics, University of Delhi — First Class
Quoting 97% accuracy for fraud detection tells a reviewer the base rate was 3%. Use precision, recall or AUC-PR and state the baseline.
Kaggle as the main evidence
Competition results show modelling skill and nothing about problem framing or deployment. Useful as a supporting line, weak as the centrepiece.
No deployment story
If none of your models reached production, say what stopped it - often the honest answer is organisational, and reviewers respect that more than vagueness.
Listing every algorithm ever studied
Twenty algorithm names in a skills block reads as coursework. Three used in anger, with the problem attached, reads as experience.
ATS considerations
Applicant tracking systems behave differently by sector, and generic advice is often wrong for a given field. These points are specific to data scientist applications:
Include "machine learning" as a phrase as well as the library names; adverts filter on the phrase.
Name the model families you have used, not just the libraries: "gradient boosting", "logistic regression", "time series forecasting".
Write "A/B testing" and "experimentation" both - different employers use different vocabulary for the same work.
Spell out degree classifications and thesis titles in plain text; research-heavy employers read them.
The Minimal ATS layout is built for this, and the ATS guide covers what parsers do to a file in more detail.
Questions about data scientist CVs
Do I need a PhD?
Not for most industry roles. It matters for research positions and for some quantitative finance and pharmaceutical work. An MSc plus production experience is the more common profile.
How should I present Kaggle results?
One line with the competition, your placing and the number of teams. Treat it as supporting evidence of modelling skill, not as your main experience.
What separates a data scientist CV from a data analyst CV?
Prediction and inference under uncertainty rather than description. If your work is mostly reporting and dashboards, the analyst framing will get better responses.
Should I list deep learning if I have only used it once?
List it under the project where you used it rather than in the headline skills. Overstating deep learning is one of the fastest ways to lose an interview.