Cross-functional Prompt Engineer at Anthropic

Role Overview
As a Cross-functional Prompt Engineer, you will own Claude’s behavioral identity across the entire product suite (claude.ai, Claude Code, and API). This role sits at the critical junction of Research and Product. You won't just be writing prompts; you will be architecting the "System Prompts" that define Claude’s personality, safety guardrails, and operational logic. You will also lead incident responses when the model displays undesirable behaviors, such as sycophancy or safety bypasses.

Key Responsibilities
System Prompt Ownership: Author and maintain the core behavior prompts for every new Claude model release.

Research Pipelines: Deliver complex "meta-prompts" for synthetic data generation, which are used to train the model via RLHF (Reinforcement Learning from Human Feedback) and Constitutional AI.

Incident Response: Lead the triage and resolution of behavioral issues, such as policy violations or reputational risks.

Behavioral Evaluations: In collaboration with the Alignment team, develop Python-based evaluations to measure how changes in prompts affect model behavior at scale.

Cross-functional Leadership: Act as the primary consultant for product teams on difficult prompting challenges.

Qualifications & Skills
Prompt Mastery: Extensive experience in multi-step prompt engineering and evaluation.

Technical Foundation: Proficient enough in Python to build evaluation frameworks from scratch; understanding of agent scaffolds and model training processes.

Soft Skills: Strong product management skills (prioritization, stakeholder management) and a background or interest in philosophy, ethics, or psychology.

Judgment: High-level intuition for distinguishing between issues that can be fixed with prompts versus those requiring model-layer retraining.

Workforce Diversity & Statistics
Anthropic is a Public Benefit Corporation (PBC) that emphasizes the necessity of diverse perspectives in building safe AI. While Anthropic does not always publish a standalone annual diversity report like legacy tech firms, the AI industry at large and Anthropic's public disclosures reflect the following approximate demographics for technical roles:

Technical Workforce Representation (Approximate AI Industry Standards 2024-2025):

Women: ~20-25%

Asian: ~35-40%

White: ~40-45%

Black / African American: ~4-6%

Hispanic / Latinx: ~6-8%

Anthropic specifically notes in this job description that underrepresented groups are more likely to experience "imposter syndrome" and urges candidates from these backgrounds to apply even if they don't meet every qualification.

Application Requirements
This application is uniquely demanding and requires a technical submission:

Written Essay: "Why Anthropic?" (200-400 words).

Behavioral Fix: Identify a specific behavior in Claude that needs fixing (2-3 sentences).

CSV Template: Provide a public spreadsheet link with one prompt that fixes the behavior and three test cases to prove its efficacy.