Skills: Python, R, SQL, NumPy, Pandas, Scikit-learn,
XGBoost, LightGBM, TensorFlow, Keras, PyTorch, HuggingFace,
Spark, MLflow, Airflow, Docker, Kubernetes, AWS, GCP, Git.

Core skills:
• Python (5 yrs) — production ML services, not just notebooks
• Scikit-learn + XGBoost — 3 shipped models (fraud, churn,
  demand forecast)
• SQL — complex joins, window functions on 100M+ row tables
• MLflow — model registry across 12 models, retrained weekly

- Achieved 0.87 AUC on a fraud detection model.

- Fraud model: 0.87 AUC vs 0.71 baseline rule engine
  (measured on 6 months of backtested production data).
  At 92% recall, false positive rate fell from 1.1% to
  0.3%, cutting manual review cost by 43% (~$28k/qtr).

- Built a churn model in a Jupyter notebook.

- Shipped churn model as a FastAPI endpoint (p95 < 80ms);
  nightly batch retraining via Airflow; MLflow registry
  tracks all 14 versions; used by the retention team daily.

Objective:
Seeking a data science role where I can apply machine
learning to impactful business problems.

Data Scientist · 4 yrs
Python · SQL · XGBoost · MLflow · FastAPI
Shipped 3 production models; fraud model cut manual
review queue 43% (~$28k/qtr saved).

Data Scientist — Nimbus Labs (2021-2024)
- Built ML models using Python and Scikit-learn.
- Worked on churn prediction and fraud detection.
- Used Pandas for data cleaning.
- Created dashboards in Power BI.
- Worked with product and marketing teams.

Projects
- Titanic survival prediction (Kaggle).
- MNIST digit classifier.

Data Scientist — Nimbus Labs (2021-2024)
- Shipped a fraud detection model to production (FastAPI,
  p95 < 80ms); 0.87 AUC vs 0.71 baseline rule engine;
  reduced manual review queue 43% (~$28k/qtr saved).
- Built a churn model for a 40k-account SaaS; top-decile
  segment (n=1.2k) targeted with offers; 2.1 pp churn
  reduction in Q3, validated against a 12k holdout.
- Designed feature pipelines (SQL + Pandas) on 100M+ row
  tables; cut training data prep time from 6 h to 40 min.
- Set up MLflow registry and weekly retraining; 14 model
  versions tracked across 3 production systems.
- Presented to VP of Growth and 2 PMs quarterly; 2 of 4
  recommendations adopted in the next sprint.

Projects
- Hospital no-show risk model (public dataset) — top-decile
  segment had 3.2× baseline no-show rate; wrote a 2-page
  intervention memo. Public notebook + repo.

Content checklist:
☐ 3-5 core skills, each with project context.
☐ Every project starts with a business problem.
☐ Metrics paired with a baseline (model vs baseline).
☐ SQL work named (joins, dedup, tuning, pipelines).
☐ No bullet copied from a job description.

Artefacts checklist:
☐ At least one project with a deployment line.
☐ No Kaggle tutorial as a primary project.
☐ One GitHub repo with README, src/, tests, and run command.
☐ One visible artefact: notebook, demo URL, or memo.
☐ One page for 2-5 years of experience.

Positioning checklist:
☐ Opening line = role + stack + one impact number.
☐ Stakeholders named (VP, PMs, product teams).
☐ Decisions moved (what changed because of the model).
☐ Every keyword defensible live.
☐ Three proofreading passes done.
☐ Filename: Firstname-Lastname-DS-Resume.pdf