French Data Labeling Associate - California based

Are you a native French speaker with a professional background in writing and a keen eye for AI safety? Welo Data is looking for Data Labeling Associates for Project Perseus.

This is not a traditional annotation role. You will be working at the intersection of human judgment and machine learning, evaluating how cutting-edge AI systems handle French language nuances, cultural context, and safety boundaries.

• Model Evaluation: Critique French AI outputs for accuracy, fluency, and tone, providing structured feedback to drive model improvement.
• Safety Engineering: Use your 2+ years of GenAI safety experience to identify bias, risks, and failure modes in Arabic datasets.
• Guideline Evolution: Apply complex linguistic guidelines thoughtfully and help refine evaluation frameworks to reflect real-world French dialects and contexts.
• Strategic Insights: Identify trends in model performance and communicate findings clearly to technical stakeholders.
• Collaborative Calibration: Participate in team syncs to ensure consistent quality across global AI workstreams.

• Employment Type: W2 Full-Time Employee
• Hours: 40 hours/week
• Location: California On-Site (San Francisco, Sunnyvale, or Burlingame)
• Pay Rate: 34 $ (per hour)

• Linguistic Mastery: Native-level French proficiency and a University Degree (Bachelor’s or higher).
• English Proficiency: B2 level or superior English skills to navigate technical guidelines and stakeholder syncs.
• Professional Writing: 1–2 years of experience in professional writing, journalism, or content creation.
• AI Safety Expertise: 2+ years of experience focusing on Safety & Trust in Generative AI data delivery.
• Critical Thinking: The ability to make sound judgment calls in ambiguous "grey area" scenarios.
• Technical Agility: A natural curiosity about LLMs and the confidence to flag issues when a model's behavior feels "off."

This is an opportunity to move beyond traditional data work and play a direct role in how AI systems are evaluated and improved. The work is fast-moving, collaborative, and increasingly central to how modern AI systems are built and deployed.