Hiring nowAlignment and scalable oversight

Model Behaviour Researcher

Also called AI Behaviour Scientist

A Model Behaviour Researcher studies how a model's behaviour shifts across contexts and training, and shows what caused the shift. That gives the company evidence it can act on when deciding how to train, evaluate and supervise its models.

What they do

They design experiments that probe how a model behaves under different prompts, settings and training choices. They run the same study across checkpoints and training runs to see which changes persist and which fade. They build adversarial evaluations to find where behaviour breaks. They write interventions that others can rerun, then test whether an effect is caused by the intervention or only observed alongside it. Much of the week goes on cleaning up experiment setups, checking results hold on a second run, and writing findings that engineers and safety colleagues can use.

What a strong hire delivers

In the first months, a strong hire delivers a behaviour study with interventions that a colleague can reproduce from the write-up alone. They set out which effects are causal and which are only observed, and say plainly how confident the team should be. They also leave behind reusable evaluation sets and experiment tooling, so the next study starts faster and the team argues less about what the results mean.

How you can tell they are good

Ask them to walk through one behaviour study from start to finish. Listen for how they chose controls, how they checked the effect replicated, and what they would not claim. Ask what they did when a result vanished on rerun. Good candidates are comfortable saying a finding is only observational. Be wary of impressive demos with no stated limits.

Why it is hard to fill

The title varies widely, so searching on it misses most of the good people. The skill sits between ML research, preference learning and technical safety, and few people have all three. Many candidates are strong at building models but weak at experimental design. Others reason well about causes but have never worked inside real training runs. Those who have both are already busy, and usually settled in research teams.

Where to find them

Look in ML research groups, preference learning and reward modelling teams, and technical safety or alignment teams. Evaluation and red-teaming engineers are another source, as are academic researchers from experimental psychology or statistics who moved into ML. They tend to sit in frontier labs, safety-focused research organisations, university labs and AI companies that train their own models.

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 Model Behaviour Researcher

  1. Describe a behaviour change you found in a model across training. How did you establish what caused it?

    A strong answer names the comparison they ran, the controls and the alternatives they ruled out. It separates cause from correlation clearly. Worry if they only describe the observation, or credit one cause without having tested it.

  2. How do you design an experiment so someone else can reproduce your intervention?

    Good answers cover fixed seeds and configurations, versioned data and prompts, written procedures and a rerun by a second person. Worry if reproducibility is treated as documentation done at the end.

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

Job description you can post

We train and deploy models, and we need to understand how their behaviour changes across contexts and training. You will run careful studies that show what changed and why, and give the team evidence it can trust when making decisions on training, evaluation and supervision.

What you will do

  • Design and run experiments on how model behaviour shifts across contexts and training stages
  • Build reproducible interventions and document them so others can rerun them
  • Separate causal findings from observational ones and state the limits clearly
  • Develop adversarial evaluations to find where behaviour breaks
  • Work with training and safety colleagues to turn findings into changes
  • Maintain experiment tooling and reusable evaluation sets

What you will bring

  • Demonstrated experimental design work on model behaviour, shown through studies or write-ups
  • Hands-on experience with preference learning or related training methods
  • Track record of evaluating models and checking that results replicate
  • Familiarity with alignment or technical safety research questions
  • Clear written communication of findings and uncertainty
  • Comfort working in training environments and experiment tooling

Nice to have

  • Background in statistics or experimental science
  • Experience with red-teaming or adversarial testing
  • Published or shared research on model behaviour

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

Skills that matter

  • Alignment researchAdvancedStudy methods to make behaviour meet intended objectives
  • Preference learningAdvancedLearn behaviour from comparisons and reward signals
  • Model evaluationWorkingMeasure capability and failure modes credibly
  • Experimental designWorkingChoose comparisons that identify credible effects

Other roles in alignment and scalable oversight

Hiring nowScalable Oversight ResearcherA scalable oversight researcher designs and tests ways to supervise AI systems on tasks too hard for a person to judge easily. The work shows…EmergingReward Hacking ResearcherA Reward Hacking Researcher finds and studies cases where models exploit the reward or evaluation signal instead of doing the intended task, then…Hiring nowModel Behaviour and Character DesignerA Model Behaviour and Character Designer defines and tunes how a model reasons, communicates and handles values trade-offs. They make the product's…

Related roles in other families

Hiring nowPost-Training Research EngineerPost-training and preference optimisationA Post-Training Research Engineer adapts pretrained models into ones that behave as the company needs. They turn preference data and reward signals…Hiring nowReward Modelling ResearcherPost-training and preference optimisationA reward modelling researcher builds models that learn what good output looks like from preference comparisons. They give post-training teams a…Hiring nowReinforcement Learning Research EngineerPost-training and preference optimisationA Reinforcement Learning Research Engineer improves how a model behaves by training it against interactive signals. They matter because reward gains…Hiring nowModel Evaluation EngineerCapability evaluation and benchmarkingA Model Evaluation Engineer builds the tests that show what an AI system can do and where it breaks. Without them, a team cannot tell whether a…

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.

← Alignment and scalable oversightAll new job families

Start a search

Hiring a Model Behaviour Researcher?

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.