AI Safety Risk Analyst: Frontier Models

BYTEDANCE PTE. LTD.; Singapore; On-site; SGD 90,000 - 150,000; Full time

Job summary

ByteDance is seeking candidates for the AI Data and Safety team to advance Seed foundational models and improve AI-native applications. You will translate ambiguous safety concerns into testable problems, work with researchers and product managers, and help build scalable data processes for rapid model iteration.

Ideal applicants bring 2+ years in analytics or research, strong problem-solving and measurement skills, and the ability to operate across data, experiments, and evaluation.

Qualifications

• Minimum 2 years in analytical, research, strategy, evaluation, or similar problem-solving role.
• Demonstrated end-to-end ownership of an ambiguous problem with measurable outputs.
• Hands-on experience using raw data, cases, or experiments as evidence.
• Strong written and verbal reasoning with clear conclusions and rationale.

Responsibilities

• Research emerging AI safety risks with industry sources, real-world cases, and expert input.
• Define risk mechanisms, scope, boundaries, and representative failure modes.
• Translate risk definitions into evaluation cases, benchmarks, metrics, and thresholds.
• Run exploratory analyses, pilots, and lightweight experiments.
• Review results, identify failure patterns, and iterate methodology.

Skills

Analytical thinking

End-to-end ownership

Research & data analysis

Written & verbal communication

SQL

Python

AI safety awareness

Cross-functional collaboration

Tools

SQL

Python

Agents

Job description

ByteDance is seeking candidates for the AI Data and Safety team to advance Seed foundational models and improve AI-native applications. You will translate ambiguous safety concerns into testable problems, work with researchers and product managers, and help build scalable data processes for rapid model iteration.

Ideal applicants bring 2+ years in analytics or research, strong problem-solving and measurement skills, and the ability to operate across data, experiments, and evaluation.