Hiring nowInterpretability and model internals

Interpretability Research Engineer

Also called Research Engineer, Interpretability

An Interpretability Research Engineer builds the tools and experiments that let researchers look inside models. They make internals work reliable and repeatable, so the team can test ideas about how a model behaves and trust the results.

What they do

They build and maintain the infrastructure researchers use to study model internals. That means instrumenting models so activations can be captured, stored and analysed at scale. They write intervention tools that let researchers change or ablate parts of a model and see what follows. They turn one-off notebooks into experiment suites others can rerun. They sit close to researchers, hear a hypothesis, and work out the quickest sound way to test it. They also chase down odd results, checking whether the cause is the model, the method or a bug.

What a strong hire delivers

In the first months, a strong hire gives researchers a dependable way to collect and analyse activations on the models the team cares about. Existing experiments become reproducible and rerunnable by someone else. Slow or fragile analysis steps get faster and clearer. Researchers spend less time fighting tooling and more time testing hypotheses, and the team starts to trust its own findings.

How you can tell they are good

Ask them to walk through an experiment suite or analysis tool they built, and what broke along the way. Strong people explain design choices, how they checked results were real and not artefacts, and what they would change. Ask how they handled memory and scale with large activations. Be wary of candidates who describe only results, not the machinery behind them.

Why it is hard to fill

Few people combine real engineering discipline with a feel for research questions. Most come from one side or the other. Strong software engineers often lack the mechanistic grounding, and researchers often write code that does not survive reuse. Titles vary between teams, and the field is small, so the best candidates are usually already inside labs or well-known research groups and are rarely searching.

Where to find them

Look at research engineers and scientists on interpretability, safety or alignment teams at frontier labs and research nonprofits. Others come from deep learning engineering, ML infrastructure or academic labs where they built tooling for experiments. Open-source contributors to interpretability libraries are a good signal. Some come from physics or neuroscience backgrounds with strong computational habits and a taste for careful experimental work.

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 an Interpretability Research Engineer

  1. Describe a tool or pipeline you built for analysing model activations. What were the hardest design decisions?

    Covers storage, memory, batching and how researchers actually used it. Names trade-offs and what they would redo. Worry if the answer is only about findings, or if they cannot explain why the tool took the shape it did.

  2. How do you make an interpretability experiment reproducible by someone who was not there?

    Mentions fixed seeds, versioned models and data, config-driven runs, logged outputs and tests on known cases. Worry if reproducibility means a notebook and a README, or if they have never had a colleague rerun their work.

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

Job description you can post

We are building our model-internals research, and we need an engineer to make it work at scale. You will build the tools and experiment infrastructure our researchers use to understand how our models represent information and why they behave as they do. The work is hands-on, close to the research, and judged by whether our findings can be trusted.

What you will do

  • Build and maintain tooling to capture and analyse model activations at scale
  • Instrument models so researchers can inspect and intervene on internal components
  • Turn exploratory analyses into reproducible, rerunnable experiment suites
  • Work with researchers to translate hypotheses into well-designed experiments
  • Investigate unexpected results and separate real effects from tooling artefacts
  • Improve the speed and reliability of analysis pipelines
  • Document tools so others can use and extend them

What you will bring

  • Demonstrated work building research tooling or experiment infrastructure for deep learning
  • Strong Python and experience with a major deep learning framework
  • Experience instrumenting or modifying models to access internal activations
  • Sound experimental design and a habit of checking results before believing them
  • Ability to handle large data and memory constraints in analysis
  • Clear communication with researchers and engineers alike

Nice to have

  • Published or open-source work in mechanistic interpretability
  • Experience with distributed or multi-device model execution
  • Background in scientific or experimental research outside machine learning

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…EmergingModel Representation AnalystA Model Representation Analyst studies how a model's internal features change with inputs and training, and tests whether those findings hold up. It…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 an Interpretability Research Engineer?

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.