Model Safety Analyst - AI Data & Safety

Role Overview

The AI Data and Safety team at ByteDance dedicates itself to evolving foundational AI models and diverse AI products, enhancing AI-native applications built on the Seed model series. Our team's focus encompasses managing the entire data lifecycle—from developing evaluation methodologies and converting user feedback into actionable data requests, to establishing scalable systems that enhance data quality for rapid model improvements.

This multidisciplinary group is composed of experts from product management, data engineering, and data operations, collaborating to understand model behavior deeply and address complex challenges promptly. By partnering with internal researchers, industry specialists, and prominent data vendors, the team addresses advanced data challenges crucial for AI advancement and a superior user experience.

The principal mission is to ensure safe, ethical AI development by identifying, monitoring, and mitigating risks through comprehensive approaches including safety assessments, training, and policy studies. The role involves rigorous testing of Seed models—including Large Language Models (LLM), Vision Language Models (VLM), Audio, and Multimodal models—in real and edge-case scenarios to detect risks. Simultaneously, it incorporates formulating and delivering safety training and refining AI safety policies to comply with regional and global standards, ultimately facilitating responsible AI innovation.

Key Responsibilities

• Investigate the latest AI safety developments in academia and industry and proactively identify gaps in existing evaluation frameworks, proposing novel test methods for real-world and edge-case scenarios.
• Create and regularly update safety evaluation guidelines for multiple model types and formulate precise, robust metrics to measure safety performance, failure patterns, and adherence to responsible AI principles.
• Analyze safety evaluation data to uncover issues related to model training, fine-tuning, or product implementation; translate these findings into practical recommendations for product and model enhancements.
• Collaborate with diverse teams to develop scalable safety evaluation workflows, establish effective feedback mechanisms for ongoing model development and risk reduction, and maintain quality assurance for safety evaluations and training data synthesis.
• Lead complete project cycles from scoping through to delivery, coordinating resource allocation and cross-functional collaboration to meet objectives and deadlines.
• Innovate and design automated processes and methodologies to improve scalability, efficiency, and comprehensiveness in safety policy research, safety data creation, data analysis, and validating evaluation frameworks.

Exposure to Sensitive Content

This position entails potential exposure to harmful or sensitive content, including but not limited to hate speech, harassment, self-harm or suicide-related materials, violence or cruelty, and child safety concerns. Support and resilience training will be provided to ensure employee well-being.

Qualifications

Required:

• Bachelor’s degree or above in AI Policy, Public Policy, Political Science, International Relations, Law, Philosophy, or related fields.
• Excellent English communication skills, both written and oral.
• Strong analytical aptitude to interpret qualitative and quantitative data and distill clear insights.
• Demonstrable project management experience, particularly with cross-functional initiatives in dynamic environments.
• Creative problem-solving capabilities and comfort working amidst ambiguity, utilizing technological tools to enhance workflow and results.

Preferred:

• Professional background in AI safety, Trust & Safety, Risk consulting, or risk management; experience within AI companies is a plus.
• Practical expertise in developing scalable automated workflows, including prompt engineering, agent skills, and automated data generation.
• High intellectual curiosity, self-drive, attention to detail, and teamwork orientation.
• Strong interest in emerging technologies, AI’s human impact, and user behaviors, along with a passion for learning from real-world scenarios.
• Technical skills helpful for automating workflows or knowledge of generative AI technologies are advantageous but not mandatory.

About ByteDance

Established in 2012, ByteDance's mission is to inspire creativity and enrich lives globally. The company offers over a dozen products including TikTok, Lemon8, CapCut, Pico, and region-specific platforms like Toutiao, Douyin, and Xigua, facilitating creative expression and content sharing worldwide.

Why Join ByteDance

ByteDance fosters innovation that empowers authentic self-expression and global connection. Embracing a diverse, enthusiastic workforce, the company values curiosity, humility, and impact-driven work in a rapidly expanding tech environment. The "Always Day 1" mindset encourages continuous growth leading to meaningful achievements and limitless possibilities.

Diversity & Inclusion

Committed to valuing diverse skills, experiences, and viewpoints, ByteDance strives to create a workplace reflecting the global communities it serves. The company champions inclusion as a core part of its mission to inspire creativity and enrich life, welcoming passionate individuals who share this vision.