German AI Prompt Evaluator — Paris, 25h/Week (Temp)

TransPerfect Translations International; Paris; Hybrid; EUR 34,000 - 55,000; Part time

Job summary

DataForce by TransPerfect seeks a detail-oriented professional to design and evaluate prompts for AI language models in Paris. The role focuses on creating high-quality prompts, defining scoring criteria, and reviewing model outputs for accuracy and compliance.

You will identify failure cases, provide actionable feedback, and collaborate with stakeholders to improve system performance. The position is on-site, 3 months, part-time (25 hours/week), starting November 2026, with potential extension.

Qualifications

• Idiomatic fluency in German.
• Advanced level of English.
• Be in Paris onsite.
• Strong attention to detail and critical thinking skills.
• Basic programming knowledge is a plus.

Responsibilities

• Develop user prompts and establish robust evaluation criteria for AI model testing.
• Review and rate model responses, ensuring compliance with project guidelines and quality standards.
• Investigate and document failure cases, providing actionable insights for improvement.
• Communicate observations, feedback, and rating justifications effectively to stakeholders.
• Apply guidelines rigorously to maintain excellent data quality.
• Identify and source German-language audio content from online platforms.
• Transcribe German audio recordings to create high-quality datasets for language model training.
• Annotate and label audio and text data according to project specifications.
• Evaluate the performance of Speech Recognition and Text-to-Speech systems.
• Curate and maintain golden datasets for model development.
• Operate robotic equipment to perform tasks and/or record and annotate video sessions.

Skills

German

Advanced English

Job description

DataForce by TransPerfect seeks a detail-oriented professional to design and evaluate prompts for AI language models in Paris. The role focuses on creating high-quality prompts, defining scoring criteria, and reviewing model outputs for accuracy and compliance.

You will identify failure cases, provide actionable feedback, and collaborate with stakeholders to improve system performance. The position is on-site, 3 months, part-time (25 hours/week), starting November 2026, with potential extension.