Data Scientist Jobs

Data scientist jobs on carbon.talent, spanning experimentation, forecasting, machine learning and decision science, in full-time, contract and fractional roles. 'Data scientist' means different things at different companies: at one it is mostly SQL and A/B tests, at another it is building production models. This guide explains how to tell which kind of role you are looking at, what interviews test, and how to present work that shows judgement rather than just modelling.

Open roles

Data science roles fall roughly into three groups. Product or decision science roles focus on experiments, causal questions and metrics, and look a lot like senior analytics with stronger statistics. Modelling roles build predictive systems such as churn, demand or fraud models, often handing them to engineers to deploy. Research-heavy roles push methods forward and are concentrated in larger companies. The listing's verbs give it away: 'design experiments' points one way, 'deploy models' another.

Core skills are consistent: Python or R, SQL, statistics including hypothesis testing and regression, and practical machine learning with libraries such as scikit-learn, XGBoost or LightGBM. Modelling roles add feature engineering, evaluation and some exposure to deployment. Many listings now mention large language models; for most data scientists that means evaluating LLM outputs, building retrieval or classification pipelines, and knowing when a simpler model would do.

Interviews test judgement more than algorithms. Expect to frame a vague business question, choose a baseline, design a fair experiment, spot data leakage and explain a result to someone without a technical background. You may get a take-home with a dataset and a deadline. The best submissions are short, start with the answer, show validation choices and admit limitations.

Your portfolio should show decisions, not just accuracy scores. A write-up that explains why you chose a time-based split, which metric matched the business cost of errors and what the model changed is more persuasive than a leaderboard position. If your best work is confidential, describe the problem, method and outcome in general terms and offer to discuss it in interview.

Data science is offered on contract more often for well-scoped modelling work, such as building a forecasting model or an experimentation framework, and fractionally by experienced scientists who set up a company's first analytics and experimentation practice. Remote data science roles are common, though teams that run many experiments may want overlap with product managers.

Career routes include senior and staff data scientist, machine learning engineering for those who enjoy production systems, analytics or data science management, and specialist paths such as causal inference or forecasting. Wherever you head, practise explaining your work in two minutes to a non-specialist. Hiring panels for senior roles test that skill constantly.

Watch for mismatches between title and work. Some data scientist listings are reporting roles with a modern title; others expect machine learning engineering. Ask in the first call what a typical month looks like and which decisions your work has changed recently. The answer tells you whether the role matches what you want to grow into.

Questions to ask before accepting: how are models deployed and who maintains them; how many experiments run each quarter and who decides what ships; what data infrastructure exists, and how much of your time will go to cleaning data versus modelling; and which recent decision was changed by data science. Teams that cannot answer the last question may not yet know how to use a data scientist well.

Applications go wrong when candidates lead with algorithms. A line such as 'built a gradient-boosted churn model with an AUC of 0.89' tells the reader little on its own. Say what decision the model supported, how you chose the threshold and what changed in the business. Hiring managers read many CVs with impressive metrics and no context.

If you are moving into data science from analysis, emphasise experiments you designed or evaluated and any forecasting or modelling you did, even if simple. If you are coming from academia, translate research into business terms and show you can deliver a useful answer quickly rather than a perfect one slowly.

Questions to ask before accepting: how are models deployed and who maintains them; how many experiments run each quarter and who decides what ships; what data infrastructure exists, and how much of your time will go to cleaning data versus modelling; and which recent decision was changed by data science. Teams that cannot answer the last question may not yet know how to use a data scientist well.

Applications go wrong when candidates lead with algorithms. A line such as 'built a gradient-boosted churn model with an AUC of 0.89' tells the reader little on its own. Say what decision the model supported, how you chose the threshold and what changed in the business. Hiring managers read many CVs with impressive metrics and no context.

If you are moving into data science from analysis, emphasise experiments you designed or evaluated and any forecasting or modelling you did, even if simple. If you are coming from academia, translate research into business terms and show you can deliver a useful answer quickly rather than a perfect one slowly.

carbon.talent lists data scientist roles with their employment type and where they are based or open to. Building a profile and applying are free.

How carbon.talent works for candidates

  • Free signup, profile and applying.
  • Contract, fractional and full-time data and AI roles.
  • carbon.talent takes no cut of a contractor's rate.
Create your free profile

Join the Data Scientist talent pool

Questions

What does a data scientist do day to day?

It depends on the role: designing and analysing experiments, building predictive models, or researching methods. Most roles involve SQL, Python or R, statistics and explaining results to non-technical colleagues.

What is tested in data scientist interviews?

Problem framing, experiment design, statistics, model evaluation, leakage and communication, often through a case discussion or take-home.

Do data scientists need to know LLMs?

Many listings mention them. For most roles that means evaluating LLM outputs and building simple pipelines around them, plus judging when a simpler approach is better.

Are data scientist roles available on contract?

Yes, especially well-scoped modelling or experimentation projects. Experienced scientists also take fractional roles setting up a company's first data science practice.