Hiring nowDataset curation and stewardship

Training Dataset Curator

Also called AI Data Curator

A Training Dataset Curator chooses, documents and maintains the data that models learn from. They make sure each dataset fits the model's goals, respects its rights and restrictions, and can be trusted and reused.

What they do

They decide what goes into a training mixture and what stays out. They write inclusion criteria, then sample, filter and deduplicate the sources to match. They check provenance and licence terms for each source and record consent where it applies. They look for overlap with evaluation sets so results are not inflated. They work with research and engineering teams to understand what a model needs, and with legal or governance colleagues on restrictions. Each release gets documentation, a catalogue entry and a change log. When a source is withdrawn or a problem appears, they trace its effect and fix it.

What a strong hire delivers

In the first months, a strong hire brings order to what exists. They produce an inventory of current datasets with known origins and rights. They ship a first dataset release with clear inclusion criteria and documentation that others can follow. They set up a repeatable check for duplicates and contamination. They agree with research how coverage and usefulness will be judged, and they give legal a reliable record of consent and licence terms.

How you can tell they are good

Ask them to walk through a dataset they built from first principles. Listen for why each source was in or out, how they measured coverage, and what they did about duplicates and test overlap. Ask to see sample documentation. Strong people can explain trade-offs plainly and name what they got wrong. Be wary of anyone who talks only about volume or treats licensing as someone else's job.

Why it is hard to fill

The title is new and rarely appears on a CV. The people who do this work are often called data scientists, archivists or governance analysts. Few have all three strands: statistical judgement, rights and consent knowledge, and hands-on filtering at scale. Most candidates are strong in one and thin in the others. We spend our time finding the mixed backgrounds and testing which strand is weakest.

Where to find them

Look to data scientists who have worked on data quality or evaluation, and to information scientists and archivists who have managed collections with rights attached. Data governance and privacy teams are another source, as are research engineers who got pulled into dataset work. They sit in model labs, AI product companies, publishers, libraries, research institutions and regulated industries that already manage consent and provenance.

Indicative pay

Not enough published evidence yet to give a fair range for this role, which is common for work this new. We benchmark it for you as part of a search.

Interview questions for a Training Dataset Curator

  1. Describe a dataset you assembled. How did you decide what to include and what to leave out?

    A strong answer states the model goal, the criteria used and the sources rejected, with reasons. It mentions rights and quality, not just size. Worry if inclusion was ad hoc, undocumented or driven by whatever was easiest to get.

  2. How have you found and handled duplicates or near-duplicates in a large collection?

    Good answers cover exact and fuzzy matching, thresholds, spot checks and the cost of over-removal. They explain how they tested the result. Worry if they only ran one tool and trusted the output without inspection.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We are looking for a Training Dataset Curator to own the data our models learn from. You will decide what goes into our training mixtures, make sure we have the rights to use it, and document each release so others can rely on it.

What you will do

  • Select and balance data mixtures to fit model goals and restrictions
  • Write and apply clear inclusion criteria for each dataset release
  • Sample, filter and deduplicate sources and check quality
  • Check for overlap between training and evaluation data
  • Record provenance, licence terms and consent in our catalogues and registers
  • Publish documentation and change logs with every release
  • Work with research, engineering and legal on data needs and restrictions

What you will bring

  • Built or maintained datasets used for training or evaluation, with examples you can discuss
  • Hands-on experience with sampling, filtering and deduplication at scale
  • Sound grasp of data provenance and quality checks
  • Working knowledge of licensing and intellectual property issues in data use
  • Experience with consent and data stewardship practices
  • Clear writing, especially documentation others have used

Nice to have

  • Background in archives, information science or data governance
  • Experience designing experiments to test data choices
  • Familiarity with dataset catalogues and consent registers

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Skills that matter

  • Data quality and provenanceAdvancedTrack dataset origin, validity and transformations
  • Data stewardship and consentAdvancedMaintain permissions, retention and responsible data access
  • Experimental designWorkingChoose comparisons that identify credible effects
  • Intellectual property and licensingWorkingAssess rights in training inputs and generated outputs

Other roles in dataset curation and stewardship

Hiring nowData Provenance SpecialistA Data Provenance Specialist traces where every dataset record came from and what was done to it, so the company can show it has the right to use…Hiring nowAI Data StewardAn AI Data Steward controls who can use each dataset, for what, and for how long. That keeps training and evaluation data within its rights, and…EmergingDataset Contamination AnalystA Dataset Contamination Analyst finds where training data overlaps with evaluation sets and recommends fixes. It matters because inflated scores…Hiring nowTraining Data Strategy LeadThe Training Data Strategy Lead decides what data the company buys, licenses, generates or collects, and manages the vendors and rights behind it…

Related roles in other families

Hiring nowPretraining Research EngineerFoundation model pretrainingA pretraining research engineer runs controlled experiments on how large models learn from broad data. The work decides which architectures, data…Hiring nowModel Architecture ResearcherFoundation model pretrainingA Model Architecture Researcher tests changes to how a model is built, to improve its behaviour or efficiency. The work tells a company which…EmergingScaling Science ResearcherFoundation model pretrainingA Scaling Science Researcher works out how model performance changes with compute, data and model size. Their estimates tell a company how to spend…EmergingTraining Dynamics ResearcherFoundation model pretrainingA Training Dynamics Researcher works out why large training runs wobble, stall or diverge, and fixes the cause. Fewer failed runs means less wasted…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert.

← Dataset curation and stewardshipAll new job families

Start a search

Hiring a Training Dataset Curator?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.