EmergingInterpretability and model internals

Model Representation Analyst

Also called Representation Analysis Researcher

A Model Representation Analyst studies how a model's internal features change with inputs and training, and tests whether those findings hold up. It matters because teams need explanations they can trust, not patterns that fall apart under pressure.

What they do

They run analyses on a model's internal activations to see what information is encoded, where, and how it shifts across inputs and training checkpoints. They design probes, then try to break their own results with sensitivity tests and interventions. They instrument models, write the analysis code, and keep experiments reproducible. They separate features that do causal work from ones that merely correlate with something nearby. They write up what they found, how confident they are, and what would change their mind, in a form researchers and engineers can act on.

What a strong hire delivers

In the first months, a strong hire delivers a representation analysis on a model the team cares about, with sensitivity tests that show which findings are stable. They set up reusable instrumentation and analysis workflows. They give the team a clear view of which explanations are trustworthy, and they retire a few that were artefacts of the method.

How you can tell they are good

Ask them to walk through a finding they later found was wrong or fragile. Strong people explain how they spotted it and what test exposed it. Ask how they would show a probe reads a feature the model uses, rather than one it merely contains. Be wary of anyone who treats probe accuracy as proof.

Why it is hard to fill

The title barely exists. This work usually sits inside research scientist, interpretability or evaluation roles, so searching by title finds almost nobody. The skill is a blend of deep learning fluency, experimental discipline and mechanistic thinking, and few people have all three. Many candidates are strong on one and thin on validation. We screen for the validation habit first.

Where to find them

Look in interpretability and mechanistic analysis groups, applied research teams at frontier labs and model providers, and academic labs working on model internals. Others come from deep learning research, evaluation, or experimental fields such as computational neuroscience, where testing claims against confounds is routine. Safety-focused organisations are also a good source.

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 Representation Analyst

  1. Describe a representation you studied. What did you conclude, and how did you check it was not a correlational artefact?

    A strong answer names the specific controls, such as interventions, alternative explanations or shifted inputs, and says what survived. Worry if they only cite probe accuracy or cannot name a single result that failed.

  2. How would you decide whether a probe finds a feature the model uses, or one it merely encodes?

    Good answers separate decodability from causal use and propose interventions such as ablation or patching. They mention probe complexity and baselines. Worry if they treat the two as the same thing.

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

Job description you can post

We are hiring a Model Representation Analyst to help us understand what our models represent internally and why. You will study how internal features change with inputs and training, and test whether our explanations hold up. Your work will shape how we trust, debug and improve our models.

What you will do

  • Analyse internal activations to identify how features vary with inputs and training.
  • Design and run probing experiments and interventions to test causal claims.
  • Build sensitivity tests that separate stable findings from correlational artefacts.
  • Instrument models and maintain reproducible analysis workflows.
  • Write clear analyses that state evidence, confidence and limits.
  • Work with researchers and engineers to turn findings into model decisions.

What you will bring

  • Demonstrated experimental work on neural network internals, shown in code, write-ups or papers.
  • Strong deep learning fluency, including training dynamics and model evaluation.
  • Experience designing controls and sensitivity tests for your own results.
  • Working knowledge of mechanistic interpretability methods.
  • Comfort writing research-grade analysis code that others can rerun.
  • Clear written communication of uncertain results.

Nice to have

  • Experience with intervention methods such as ablation or activation patching.
  • Background in a field that tests claims against confounds, such as computational neuroscience.
  • Published or open analysis of model internals.

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

Skills that matter

  • Mechanistic interpretabilityAdvancedInvestigate internal model mechanisms
  • Deep learningAdvancedBuild and diagnose trainable neural models
  • Experimental designWorkingChoose comparisons that identify credible effects
  • Model evaluationWorkingMeasure capability and failure modes credibly

Other roles in interpretability and model internals

Hiring nowMechanistic Interpretability ResearcherA mechanistic interpretability researcher works out how a model computes its outputs, using interventions and controlled tests. It matters because a…Hiring nowInterpretability Research EngineerAn Interpretability Research Engineer builds the tools and experiments that let researchers look inside models. They make internals work reliable…EmergingInterpretability Evaluation ScientistAn Interpretability Evaluation Scientist tests whether explanations of model behaviour and discovered internal features are actually faithful. That…

Related roles in other families

Hiring nowMachine Learning Research ScientistLearning theory and statistical machine learningA Machine Learning Research Scientist develops and tests learning methods against explicit hypotheses. They give the company evidence about what…Hiring nowStatistical Learning ScientistLearning theory and statistical machine learningA Statistical Learning Scientist builds prediction methods that show their assumptions and their uncertainty. For a company shipping models into…Hiring nowLearning Theory ResearcherLearning theory and statistical machine learningA Learning Theory Researcher works out what learning algorithms can and cannot do, and proves it. For an AI company, that means firmer ground for…Hiring nowAutomated Machine Learning ResearcherLearning theory and statistical machine learningAn Automated Machine Learning Researcher builds search procedures that choose and tune models within a fixed compute budget. It matters because…

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.

← Interpretability and model internalsAll new job families

Start a search

Hiring a Model Representation Analyst?

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.