We use a small set of essential cookies / browser storage to keep you signed in and to remember which CV you're editing. With your consent we'd also like to use analytics cookies to understand how the product is used. We do not run advertising trackers. See our Cookie Policy and Privacy Policy.
Data scientist CVs are judged on business impact, not model zoo. Recruiters want the problem you framed, the method you chose, and the number that moved in production. 'Built an ML model' with no outcome reads like a bootcamp project.
Python (pandas, scikit-learn, PyTorch or TensorFlow)
SQL, still the most-used tool; name the warehouse (Snowflake, BigQuery)
Modelling: regression, gradient boosting (XGBoost/LightGBM), clustering
ML in production: feature stores, MLflow, model monitoring, drift
Experimentation: A/B testing, causal inference, statistical significance
NLP / LLMs if relevant: embeddings, RAG, fine-tuning, prompt evaluation
Cloud + orchestration: AWS SageMaker / GCP Vertex, Airflow, dbt
Only include skills you can defend in an interview. Inflated skill lists fail the first phone screen.
A business outcome per project: revenue, churn, cost, or a decision changed
Whether the model actually shipped (served in production vs notebook only)
Scale context: rows, users, predictions/day, latency budget
A link to a notebook, paper, or write-up that shows reasoning, not just accuracy
Weak
Built a machine learning model to predict customer churn.
Strong
Shipped an LightGBM churn model (AUC 0.86) served via SageMaker at 40k scores/day; the retention team's targeted offers cut monthly churn from 4.1% to 3.2%.
Weak
Did exploratory data analysis and feature engineering.
Strong
Built a 120-feature store in dbt + Feast that 3 downstream models now share; cut new-model time-to-production from 5 weeks to 8 days.
Weak
Used NLP for text classification.
Strong
Replaced a keyword ticket-router with a fine-tuned embedding classifier (macro-F1 0.91); auto-routed 78% of 12k weekly tickets, saving ~2 FTE of triage.
A Kaggle notebook that explains the why, not just the leaderboard score
A deployed model with a live demo (HuggingFace Space, Streamlit)
A blog post reproducing and critiquing a published result
An open-source contribution to a DS/ML library
Upload your existing CV, paste any data scientist job description, and CVCL produces a tailored CV + cover letter in 60 seconds. 3 free generations on signup.