Remote Java QA Engineer for AI Model Training

Outlier; United States; Remote; USD 28.000 - 47.000; Vollzeit

Benefits dieser Stelle

Free access to leading LLMs

Flexible remote work

Up to 40 hrs/week

Zusammenfassung

Outlier is recruiting QA Coders to help train AI models by reviewing AI-generated code and providing structured feedback. You can work remotely from anywhere in the world on a freelance 1099 basis, with flexible hours up to 40 per week.

You'll rate model responses, flag issues, and craft clear computer science prompts to push AI capabilities. A Bachelor's in CS and experience contributing to production codebases are preferred, with strong written and spoken English required.

Qualifikationen

• Bachelor's degree in Computer Science from a selective institution.
• Software engineering experience contributing to a live production codebase.
• Ability to articulate complex concepts fluently in written and spoken English.
• Excellent attention to detail, including grammar, punctuation, and style guidelines.

Aufgaben

• Review AI-generated code: Read through code produced by AI models and flag errors, inconsistencies, or quality issues using structured guidelines.
• Rate and rank model responses: Score responses across key dimensions - accuracy, instruction following, and presentation - to help improve model output.
• Write computer science questions: Develop clear, well-structured prompts that challenge AI models and push the boundaries of their coding ability.

Kenntnisse

Attention to detail

English proficiency

QA/code review

Ausbildung

Bachelor's degree in Computer Science

Jobbeschreibung

Outlier is recruiting QA Coders to help train AI models by reviewing AI-generated code and providing structured feedback. You can work remotely from anywhere in the world on a freelance 1099 basis, with flexible hours up to 40 per week.

You'll rate model responses, flag issues, and craft clear computer science prompts to push AI capabilities. A Bachelor's in CS and experience contributing to production codebases are preferred, with strong written and spoken English required.