Bilingual German Expert - AI Safety

Obsidian; Berlin; Vor Ort; EUR 45.000 - 60.000; Vollzeit

Zusammenfassung

Obsidian is seeking proficient bilingual speakers (English & German) to help make advanced AI models safer. You will apply your language fluency and cultural judgment to evaluate how these models handle sensitive topics in German. No prior AI/ML experience is required — we will train you on the workflow.

You will write expert-level prompts in German, apply guidelines to classify prompts and conversations, and flag adversarial phrasing while documenting the reasoning behind your judgments.

Qualifikationen

• Native or near-native German, plus business-level written English.
• A bachelor's degree, completed or in progress.
• Strong written reasoning and a high level of attention to detail.
• Sound judgment around sensitive and dual-use information.

Aufgaben

• Write expert-level prompts in German across a range of sensitive subject areas.
• Apply structured guidelines to classify prompts and conversations.
• Flag adversarial phrasings and escalation patterns, and document the reasoning behind your judgments.

Kenntnisse

German fluency

English proficiency

Ausbildung

Bachelor's degree

Jobbeschreibung

About the role

We're hiring proficient bilingual speakers (English & German) to help make advanced AI models safer. You'll apply your language fluency and cultural judgment to evaluate and strengthen how these models handle sensitive topics in German. No prior AI/ML experience is required — we'll train you on the workflow.

What you'll do

• Write expert-level prompts in German across a range of sensitive subject areas.
• Apply structured guidelines to classify prompts and conversations.
• Flag adversarial phrasings and escalation patterns, and document the reasoning behind your judgments.

What we're looking for

• Native or near-native fluency in German, plus business-level written English.
• A bachelor's degree, completed or in progress.
• Strong written reasoning and a high level of attention to detail.
• Sound judgment around sensitive and dual-use information.

Nice to have

• Based in Germany or the wider Western Europe region. This is a preference, not a requirement — applicants based elsewhere are welcome.
• Experience reviewing, grading, or red-teaming written or technical content.
• Background in trust and safety, content moderation, policy evaluation, or adversarial testing.

Project Timeline

• Start date: Immediate