AI Safety & Evaluation Research Engineer

OJ Digital; San Francisco (CA); On-site; USD 120,000 - 160,000; Full time

Benefits offered by this job

Equity

In-person with flexibility

Relocation assistance

Visa sponsorship

Job summary

OJ Digital is seeking exceptional Research Engineers to join its San Francisco team. You will work on challenging AI safety, model behavior and evaluation tasks, combining software engineering, AI research and hands-on work with advanced AI systems.

You’ll have significant ownership from day one, collaborating with a small, highly capable engineering and research team on technically difficult problems and bold approaches to ensuring safe, reliable AI behavior.

Qualifications

• 1+ years of experience in Software Engineering, ML Engineering, AI Research or related field.
• Exceptional generalist software engineering ability.
• Hands-on experience evaluating, testing, monitoring or critically analysing AI/ML systems.

Responsibilities

• Build and own complex technical projects from initial idea through to delivery.
• Design and improve methods for testing and evaluating the behaviour of advanced AI systems.
• Investigate where models or agents fail or produce unreliable outputs.
• Work with LLMs and AI agents, analysing and critically evaluating their outputs.
• Build primarily in Python and develop evaluation frameworks and verification systems.
• Identify subtle weaknesses that automated systems may miss and iteratively improve.
• Continuously test, refine and improve the quality and reliability of your work.

Skills

Software Engineering

ML Engineering

AI Research

Python

Problem solving

Job description

OJ Digital is seeking exceptional Research Engineers to join its San Francisco team. You will work on challenging AI safety, model behavior and evaluation tasks, combining software engineering, AI research and hands-on work with advanced AI systems.

You’ll have significant ownership from day one, collaborating with a small, highly capable engineering and research team on technically difficult problems and bold approaches to ensuring safe, reliable AI behavior.