Propulsion Systems Engineer — AI Safety & Dual-Use (Remote)

Energy Jobline ZR; San Francisco (CA); Remote; USD 90,000 - 103,000; Full time

Job summary

Mercor, headquartered in San Francisco, seeks a Propulsion & Effects Engineer for a contract role to craft prompts in propulsion and energetic materials and to evaluate AI responses for policy compliance.

You will write reference answers with technical reasoning and collaborate with experts to ensure AI models distinguish routine from potentially dangerous queries, working independently to improve model performance on dual-use topics.

Qualifications

• Experience in munition s or warhead design and effects engineering.
• Background in propulsion system design, test, and qualification.
• Expertise in terminal effects modelling such as blast, fragmentation, and penetration.
• Knowledge of fuzing, safe-and-arm, and initiation system engineering.
• Experience in test range engineering and instrumentation for propulsion or ordnance.
• Prior technical writing, published research, or expert witness experience.

Responsibilities

• Write challenging single-turn prompts in propulsion and energetic materials.
• Evaluate AI model responses against policy standards. Judge the correctness of each response.
• Write reference answers. Provide technical reasoning for correct responses.
• Collaborate with experts to ensure AI models distinguish between routine and potentially dangerous queries.
• Work independently to improve AI model performance in handling dual-use scenarios.

Skills

Propulsion design

Warhead design

Test and qualification

Terminal effects modelling

Fuzing & safing

Instrumentation

Technical writing

Job description

Mercor, headquartered in San Francisco, seeks a Propulsion & Effects Engineer for a contract role to craft prompts in propulsion and energetic materials and to evaluate AI responses for policy compliance.

You will write reference answers with technical reasoning and collaborate with experts to ensure AI models distinguish routine from potentially dangerous queries, working independently to improve model performance on dual-use topics.