Research Engineer, Interpretability

Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.

When you see what modern language models are capable of, do you wonder, "How do these things work? How can we trust them?"

The Interpretability team at Anthropic is working to reverse-engineer how trained models work because we believe that a mechanistic understanding is the most robust way to make advanced systems safe.

Think of us as doing "neuroscience" of neural networks using "microscopes" we build - or reverse-engineering neural networks like binary programs.

Even if you haven’t worked on interpretability before, the infrastructure expertise is similar to what's needed across the lifecycle of a production language model:

The science keeps scaling - and it's now applied directly in safety audits on frontier models, with real deadlines. As our research has matured, engineering and infrastructure have become a bottleneck. Your work will have a direct impact on one of the most important open problems in AI.

• Our research blog - covering advances including Monosemantic Features and Circuits
• An Introduction to Interpretability from our research lead, Chris Olah
• The Urgency of Interpretability from CEO Dario Amodei
• Engineering Challenges Scaling Interpretability - directly relevant to this role
• 60 Minutes segment - Around 8:07, see a demo of tooling our team built
• New Yorker article - what it's like to work on one of AI's hardest open problems
• Pretraining: Training dictionary learning models looks a lot like model pretraining - creating stable, performant training jobs for massively parameterized models across thousands of chips
• Inference: Interp runs a customized inference stack. Day-to-day analysis requires services that allow editing a model's internal activations mid-forward-pass - for example, adding a "steering vector"
• Performance: Like all LLM work, we push up against the limits of hardware and software. Rather than squeezing the last 0.1%, we are focused on finding bottlenecks, fixing them and moving ahead given rapidly evolving research and safety mission

• Build and maintain the specialized inference and training infrastructure that powers interpretability research - including instrumented forward/backward passes, activation extraction, and steering vector application
• Resolve scaling and efficiency bottlenecks through profiling, optimization, and close collaboration with peer infrastructure teams
• Design tools, abstractions, and platforms that enable researchers to rapidly experiment without hitting engineering barriers
• Help bring interpretability research into production safety audits - with real deadlines and high reliability expectations
• Work across the stack - from model internals and accelerator-level optimization to user-facing research tooling

• Have 5-10+ years of experience building software
• Are highly proficient in at least one programming language (e.g., Python, Rust, Go, Java) and productive with Python
• Are extremely curious about unfamiliar domains; can quickly learn and put that knowledge to work, e.g. diving into new layers of the stack to find bottlenecks
• Have a strong ability to prioritize the most impactful work and are comfortable operating with ambiguity and questioning assumptions
• Prefer fast-moving collaborative projects to extensive solo efforts

• Optimizing the performance of large-scale distributed systems
• Language modeling fundamentals with transformers
• High Performance LLM optimization: memory management, compute efficiency, parallelism strategies, inference throughput optimization
• Working hands-on in a mainstream ML stack - PyTorch/CUDA on GPUs or JAX/XLA on TPUs
• Collaborating closely with researchers and building tooling to support research teams; or directly performed research with complex engineering challenges

• Building Garcon, a tool that allows researchers to easily instrument LLMs to extract internal activations
• Designing and optimizing a pipeline to efficiently collect petabytes of transformer activations and shuffle them
• Profiling and optimizing ML training jobs, including multi-GPU parallelism and memory optimization
• Building a steered inference system that applies targeted interventions to model internals at scale (conceptually similar to Golden Gate Claude but for safety research)

The annual compensation range for this role is listed below.

For sales roles, the range provided is the role’s On Target Earnings ("OTE") range, meaning that the range includes both the sales commissions/sales bonuses target and annual base salary for the role.

• This role is based in the San Francisco office; however, we are open to considering exceptional candidates for remote work on a case-by-case basis.

$315,000—$560,000 USD

Logistics

Minimum education: Bachelor’s degree or an equivalent combination of education, training, and/or experience

Required field of study: A field relevant to the role as demonstrated through coursework, training, or professional experience

Minimum years of experience: Years of experience required will correlate with the internal job level requirements for the position

Location-based hybrid policy: Currently, we expect all staff to be in one of our offices at least 25% of the time. However, some roles may require more time in our offices.

Visa sponsorship: We do sponsor visas! However, we aren't able to successfully sponsor visas for every role and every candidate. But if we make you an offer, we will make every reasonable effort to get you a visa, and we retain an immigration lawyer to help with this.

We encourage you to apply even if you do not believe you meet every single qualification. Not all strong candidates will meet every single qualification as listed. Research shows that people who identify as being from underrepresented groups are more prone to experiencing imposter syndrome and doubting the strength of their candidacy, so we urge you not to exclude yourself prematurely and to submit an application if you're interested in this work. We think AI systems like the ones we're building have enormous social and ethical implications. We think this makes representation even more important, and we strive to include a range of diverse perspectives on our team.

Your safety matters to us. To protect yourself from potential scams, remember that Anthropic recruiters only contact you from @anthropic.com email addresses. In some cases, we may partner with vetted recruiting agencies who will identify themselves as working on behalf of Anthropic. Be cautious of emails from other domains. Legitimate Anthropic recruiters will never ask for money, fees, or banking information before your first day. If you're ever unsure about a communication, don't click any links—visit anthropic.com/careers directly for confirmed position openings.

How we're different

We believe that the highest-impact AI research will be big science. At Anthropic we work as a single cohesive team on just a few large-scale research efforts. And we value impact — advancing our long-term goals of steerable, trustworthy AI — rather than work on smaller and more specific puzzles. We view AI research as an empirical science, which has as much in common with physics and biology as with traditional efforts in computer science. We're an extremely collaborative group, and we host frequent research discussions to ensure that we are pursuing the highest-impact work at any given time. As such, we greatly value communication skills.