Senior Model Behavior Engineer: Reasoning & Decision Making

Elicit; Oakland (CA); Hybrid; USD 185,000 - 320,000; Full time

Benefits offered by this job

Health insurance

401K with employer match

Mac setup and home-office budget

AI learning budget

Job summary

Elicit is seeking a role focused on enhancing how teams reason with evidence. You will own evaluations, decompose problems, and improve agent-driven workflows while collaborating with engineers, product, and domain experts.

The work centers on systems built around frontier models with a strong emphasis on legible reasoning and testable improvements. You’ll design abstractions, write maintainable code in Python, and help ship reliable, inspectable decision-support tools.

Qualifications

• Ability to make vague questions precise and explain assumptions clearly.
• Ability to design experiments and evaluate results rigorously.
• Experience turning failures into testable hypotheses.
• Ability to implement ideas in maintainable code and tests.
• Comfort with Python or a comparable language and AI toolchains.

Responsibilities

• Understand the decisions our users are trying to make by collaborating with domain experts.
• Define and measure what good reasoning looks like across problems.
• Improve the system by iterating on problem decomposition and evidence gathering.
• Design software boundaries and abstractions for reliable reasoning workflows.
• Own the quality of what you ship through reviews, tests, and monitoring.

Skills

Rigorous thinking

Curiosity across domains

Experimental instincts

Language models familiarity

Engineering judgment

Effective use of coding assistants

Python programming

Tools

Python

Job description

Elicit is seeking a role focused on enhancing how teams reason with evidence. You will own evaluations, decompose problems, and improve agent-driven workflows while collaborating with engineers, product, and domain experts.

The work centers on systems built around frontier models with a strong emphasis on legible reasoning and testable improvements. You’ll design abstractions, write maintainable code in Python, and help ship reliable, inspectable decision-support tools.