Cake Job Search

About Us AIFT is at the forefront of AI safety and security, developing cutting-edge solutions, Vulcan Attack (automated vulnerability assessment/red teaming) and Vulcan Protect (real-time monitoring), to ensure AI systems are robust, trustworthy, and aligned with societal needs. Our Product Vulcan is a cybersecurity solution specifically designed for GenAI, offering two core services: Red Team (vulnerability assessment) and Blue Team (real-time defense). It ensures GenAI compliance, cybersecurity robustness, and operational integrity. Since its official launch in 2024, Vulcan has been recognized by the international standard-setting organization OWASP as a certified vendor for LLM GenAI security testing and assessment. It is one of the few solutions capable of supporting multiple Asian languages (Traditional Chinese, Simplified Chinese, Japanese, Korean, Thai) and Standard Arabic. The Role We are seeking a passionate and highly skilled GenAI Security Researcher to join our team. You will conduct applied and theoretical research into AI vulnerabilities, attack/defense methods, and security/safety benchmarking. Your work will directly shape our AI safety products, contribute to research communities, and influence industry standards and regulations. Location Asia (e.g. Taipei, Hong Kong, Singapore, South Korea, Japan) or Remote What You'll Do Research: Investigate adversarial attacks, data poisoning, prompt injection, jailbreaks, and other AI system vulnerabilities in multi-modalities (e.g., text, image, video, and audio). Design and evaluate mitigation strategies, including fine-tuning, guardrails, content filtering, and robust model architectures.Benchmarking: Develop and maintain AI security evaluation frameworks, datasets, and performance metrics.Collaboration: Partner with product and partnership teams to integrate the findings into our AI security solutions.Dissemination: Publish findings in top-tier conferences and journals, present at industry events, and contribute to open-source initiatives.