Software Engineer, Tokens and Prompt Structures

The Encodings Infra team maintains the libraries that engineers and researchers across Anthropic use to encode text and multimodal data into a form that Claude can consume. It also determines Claude’s prompt shape: how a user’s turn is represented to the model, how Claude calls tools and receives tool results, and so on.

As a Software Engineer on this team, you’ll own the design and maintenance of these libraries-keeping their APIs intuitive, their performance sharp, and their abstractions solid enough that most of the org never has to think about encodings or prompt structures at all. You’ll have the satisfaction of knowing that your work enabled Claude to learn new ways of understanding the world.

This role is unusually broad: your work will touch systems across the codebase, from pretraining to finetuning to the API, and you’ll collaborate closely with both researchers and engineers to make sure new encoding ideas can move quickly from experiment to production., * Maintain and improve the encoding libraries used by engineers and researchers across Anthropic

• Run experiments to determine the optimal way to feed structured data into Claude without confusing it
• Design data structures and abstractions that shield most of the organization from the details of how encoded data works while enabling “power users”
• Adapt the encoding libraries to support new research directions as they emerge, and make sure that we can ship these research ideas to production
• Optimize encoding performance across the systems that depend on these libraries

• Have 5+ years of software engineering experience, with meaningful time spent maintaining libraries, SDKs, or developer-facing APIs
• Have familiarity with ML terminology and LLM architecture - you don’t need to be an ML expert, but enough understanding to work effectively alongside researchers and understand the results of experiments
• Have experience carrying out complex refactors in large codebases
• Have strong communication skills and enjoy working closely with researchers and engineers to understand what they need
• Are results-oriented, with a bias towards flexibility and impact
• Pick up slack, even if it goes outside your job description
• Care about the societal impacts of your work

Strong candidates may also have experience with:

• Tokenizers or other text/data encoding systems
• Maintaining a widely-used library over a long period of time
• Performance optimization
• Python and/or Rust
• Reinforcement learning or model training infrastructure, 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.

• Working with a research team to ship a new multimodal data type (audio, video, etc) to production
• Redesigning a core abstraction so that we can change how data is encoded into Claude without breaking downstream teams

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. Annual Salary: $320,000-$405,000 USD, Anthropic is a public benefit corporation headquartered in San Francisco. We offer competitive compensation and benefits, optional equity donation matching, generous vacation and parental leave, flexible working hours, and a lovely office space in which to collaborate with colleagues. Guidance on Candidates’ AI Usage: Learn about our policy for using AI in our application process.

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., 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.

The easiest way to understand our research directions is to read our recent research. This research continues many of the directions our team worked on prior to Anthropic, including: GPT-3, Circuit-Based Interpretability, Multimodal Neurons, Scaling Laws, AI & Compute, Concrete Problems in AI Safety, and Learning from Human Preferences.