AI Safety Researcher

Obiguard; United States; Hybrid; USD 22,000 - 37,000; Full time

Job summary

Obiguard is seeking a research-focused security professional to study risks affecting large language models and AI agents, translating findings into practical safeguards for production systems. You will investigate attack techniques, build detectors, and help shape policies governing model and tool use.

Based in Kuala Lumpur, you will join a full-time role with hybrid work options, combining remote work and on-site collaboration as needed.

Qualifications

• Based in Kuala Lumpur or Klang Valley and authorized to work in Malaysia.
• Available for full-time work with hybrid arrangements for on-site visits.

Responsibilities

• Investigate emerging attacks against language models and agent-based systems.
• Develop production-ready detectors and policy controls from research findings.
• Explore risks involving prompt injection, jailbreaks, and agent tool use.
• Contribute to red-team evaluations and analysis of AI system behavior.

Skills

AI security

Red team

Threat detection

Policy development

Research methods

Tools

Agent frameworks

Job description

Role overview

Research security risks affecting large language models and AI agents, then turn findings into practical safeguards for production systems. The work spans investigating attack techniques, building detection methods, and helping shape policies that govern model and tool use.

Responsibilities

• Investigate emerging attacks against language models and agent-based systems.
• Develop production-ready detectors and policy controls from research findings.
• Explore risks involving prompt injection, jailbreaks, and agent tool use.
• Contribute to red-team evaluations and analysis of AI system behavior.

Requirements

• Be based in Kuala Lumpur or the Klang Valley and already authorized to work in Malaysia.
• Be available for full-time work, with remote work within the area and office visits when needed.

Nice to have

• Publications or public writing on AI security, adversarial machine learning, or red teaming.
• Experience with capture-the-flag events, bug bounties, or formal security research.
• Familiarity with agent development frameworks or tooling.