Data · Senior

Data Scientist Resume Example

This example suits a data scientist with five or more years of experience who builds models that run in production. It shows how to report evaluation metrics against a baseline, link them to results measured by experiment, and describe deployment work done with engineers.

The example resume

Laid out in the Abstract template. The person, employers, schools and figures in this example are invented for illustration. Any resemblance to real people or organisations is unintended.

Wei Ling Tan

Senior Data Scientist | Forecasting, Experimentation, Production ML

  • weiling.tan@example.com
  • +1 555 0167
  • Singapore
  • linkedin.example.com/in/weiling-tan
  • github.example.com/weiling-tan

Professional summary

Senior data scientist with eight years of experience building and deploying models in logistics and telecoms. Specialises in demand forecasting, uplift modelling and experiment design, and works closely with engineers to take models from notebook to monitored production service. Leads a team of three.

Work experience

Senior Data Scientist

Pasir Lane Logistics, Singapore

Jan 2022 – Present

  • Built a LightGBM delivery-time model on 60 million historical orders, reducing mean absolute error from 9.4 to 6.1 minutes against the rule-based estimate
  • Designed a hierarchical demand forecast for 140 delivery zones that cut weekly MAPE from 21% to 13% and now drives courier shift planning
  • Set up the experimentation framework for pricing and dispatch tests with product managers, covering power calculations, guardrail metrics and switchback designs for marketplace interference
  • Deployed models with ML engineers as containerised services on Kubernetes using MLflow and Airflow, adding drift monitoring and automated weekly retraining
  • Led and mentored three data scientists, introducing code review and a model card for every production model

Data Scientist

Northquay Telecom, Singapore

Jul 2018 – Dec 2021

  • Developed an XGBoost churn model (AUC 0.84) for 1.2 million prepaid subscribers; targeted offers cut monthly churn in the top-risk decile by 3.1 points in a controlled trial
  • Replaced propensity targeting with an uplift model for upgrade campaigns, raising incremental conversions by 18% at the same contact volume
  • Built a PySpark feature pipeline over call, top-up and network data, moving model refresh from monthly to weekly
  • Forecast daily data traffic for 900 cell sites to support capacity planning, improving on the previous seasonal model by 22% (MAPE) in backtests
  • Presented results and limitations to marketing leadership each quarter, including two tests that showed no effect

Skills

Modelling: gradient boosting, GLMs, time-series forecasting, uplift and causal inference, survival analysis

Experimentation: A/B and switchback tests, power analysis, CUPED variance reduction

Languages and libraries: Python (pandas, scikit-learn, LightGBM, statsmodels, PyTorch), SQL, PySpark

Production: MLflow, Airflow, Docker, Kubernetes, BigQuery, model monitoring

Education

MSc Statistics

Wexcombe University, United Kingdom

2017 – 2018

Dissertation: Bayesian hierarchical models for sparse count data

BSc Mathematics

Eastern Straits University, Singapore

2013 – 2017

Talks and writing

Switchback experiments for dispatch pricing: what we learned

Strait Data Science Conference

2024

Hierarchical forecasting for zone-level delivery demand

Pasir Lane Logistics engineering blog

2023

Open this example in the resume builder Start empty with this template

Why this resume works

  • Each model is described with its evaluation metric and a baseline: mean absolute error falling from 9.4 to 6.1 minutes, MAPE from 21% to 13%, an AUC of 0.84. A technical reader can judge the work without guessing.
  • Model metrics are linked to effects measured by experiment, such as the 3.1 point fall in churn in a controlled trial. That separates this document from one that reports offline accuracy alone.
  • Deployment is explicit. The document names MLflow, Airflow and Kubernetes, mentions drift monitoring and retraining, and credits the ML engineers, so it is clear the models ran in production.
  • The experimentation bullet uses precise terms (power calculations, guardrail metrics, switchback designs) that signal real experience of testing in a marketplace, where ordinary A/B tests can mislead.
  • Leadership is shown through practices introduced, code review and model cards, plus one honest line about presenting tests that showed no effect.

What employers look for in this role

  • Frame business problems as modelling or experimentation problems and agree success measures with stakeholders
  • Build, validate and tune statistical and machine learning models using Python and SQL
  • Design and analyse experiments, including sample size, metrics and interpretation of results
  • Work with data and ML engineers to deploy models, monitor them for drift and retrain when needed
  • Engineer features from large, messy data sets and keep pipelines reproducible
  • Explain methods, results and limitations to product, commercial and leadership audiences
  • Review the work of other data scientists and set standards for code and evaluation

Skills to consider

List only the skills you can back up with an example. These are the ones that come up most for this kind of role.

Modelling and statistics
Regression and GLMs, Gradient boosting (XGBoost, LightGBM), Time-series forecasting, Causal inference and uplift modelling, Bayesian methods, Deep learning with PyTorch or TensorFlow, NLP and embeddings
Experimentation and evaluation
A/B test design and power analysis, Variance reduction (CUPED), Cross-validation and backtesting, Model calibration, AUC, precision and recall, MAE and MAPE
Engineering and deployment
Python (pandas, scikit-learn), SQL, Spark or PySpark, MLflow or another model registry, Airflow, Docker, Cloud ML services, Git and code review
Communication and leadership
Problem framing with stakeholders, Explaining uncertainty, Technical writing, Mentoring and code review

More bullet points to adapt

Each of these follows the same pattern as the example: an action, what it was applied to, and what changed. Swap in your own facts and figures. The bullet builder assembles the pattern for you.

  • Built a two-stage recommender (candidate retrieval plus a ranking model) that lifted click-through by 11% in a four-week A/B test on 2.3 million users
  • Applied CUPED variance reduction on the experimentation platform, cutting the sample needed to detect a 1% change in conversion by roughly a third
  • Calibrated a credit default model with isotonic regression and showed stable performance across four customer segments before model risk sign-off
  • Replaced a nightly batch scoring job with a real-time inference service at a p95 latency of 45 ms, working with two backend engineers
  • Detected feature drift after a pricing change using population stability index checks and retrained the model within the same week
  • Wrote the offline evaluation suite for a fraud model, reporting precision at fixed recall and expected cost per thousand transactions instead of accuracy

Words that recur in adverts

These terms appear often in adverts for this kind of role. Use the ones that are true of you, in sentences that show the skill in use, and check a specific advert with the keyword checker.

  • Machine learning
  • Python
  • SQL
  • Statistical modelling
  • Experimentation
  • A/B testing
  • Causal inference
  • Forecasting
  • Feature engineering
  • Model deployment
  • MLOps
  • Model monitoring
  • Deep learning
  • Stakeholder communication

Which sections to include

Open with a summary that states your specialism (forecasting, recommendation, NLP, experimentation) and whether your models run in production. Experience carries the page: for each role, lead with the model or experiment that had the largest measured effect and include its evaluation metric. Place a grouped skills section after experience, separating modelling, experimentation and engineering tools so that a data science lead and a recruiter can both find what they need. Education counts for more here than in many fields, so give degree subjects and a dissertation title if it is relevant. Publications or talks go last, one line each.

Mistakes to avoid

  • Reporting accuracy with no baseline or context. State the metric that suits the problem, what it was before your model and, where possible, the business result measured in a test.
  • Listing every algorithm you have studied. Readers want the methods you have applied to real problems, and a long list of techniques with no project behind them weakens the page.
  • Leaving it unclear whether a model was ever used. Say whether it reached production, how it was served and monitored, and who you worked with to get it there.
  • Presenting competition or coursework scores as business impact. They are fine as supporting detail, but they are different from a model that changed a decision.
  • Hiding the statistics. Experiment design, causal reasoning and honest handling of null results distinguish a senior data scientist, so give them at least one bullet.

Questions and answers

Which metrics should I put on a data scientist resume?

Use the metric your team used to judge the model, with its baseline: AUC or precision at a given recall for classification, MAE or MAPE for forecasts, incremental conversions for targeting. Then add the business measure if an experiment confirmed it. Avoid bare accuracy on imbalanced problems, since an informed reader will discount it.

Should I include publications, talks or competition results?

Include them if they are recent and relevant to the role, in a short section near the end. A conference talk or paper supports a senior or research-leaning application. Keep each to one line, and do not let the section push production work off the page.

How do I describe a model that never reached production?

Say so plainly and state what the work delivered instead: a finding that changed a decision, a prototype handed to engineering, or evidence that the idea was not worth building. Many projects end this way, and a clear account of what was learned reads better than an implied launch.

How is this different from a data analyst resume?

A data scientist resume centres on models and experiments: how they were built, evaluated, deployed and monitored. An analyst resume centres on reporting, dashboards and answering questions with existing data. The data analyst example shows that framing; choose the one that matches the advert, whatever your current job title says.