Research Engineer I (Computer Science / Artificial Intelligence / Cyber Security)

The Alibaba-NTU Global e-Sustainability CorpLab (ANGEL) represents a key collaboration between Alibaba Group and Nanyang Technological University (NTU). Supported by the Singapore RIE2025 Fund, ANGEL creates and deploys impactful green digital technologies for global sustainability. ANGEL’s mission aligns seamlessly with Singapore’s national objectives, such as the Singapore Green Plan 2030 and the Singapore Smart Nation Initiative. ANGEL’s work in developing sustainable and equitable digital solutions and promoting a sustainable and green lifestyle will contribute to a smaller carbon footprint. Its holistic approach will help to secure a brighter, more sustainable future for humanity. ANGEL will also cultivate human talents equipped with both advanced technical skills and a solid understanding of sustainable practices. They will enable Singapore to realise its environmental and technological ambitions. We are seeking to appoint a Research Engineer who will contribute to NTU’s mission of advancing trustworthy AI and cyber security research, with a particular focus on the runtime safety of autonomous coding agents. The Research Assistant will support ongoing research on safe reinforcement learning for coding agents operating in long-horizon, executable repository-and-shell environments, assisting in the construction of RL training datasets, the implementation of safety-alignment training, and the evaluation of agent safety and effectiveness. The role will contribute to NTU’s strategic initiatives in building secure, reliable, and resilient AI agent ecosystems, while supporting collaborations across academia and industry under the guidance of senior researchers. Key responsibilities: Assist in conducting research on safety alignment of executable coding agents using reinforcement learning in long-horizon repository-and-shell environments. Support the construction of RL training datasets by collecting, synthesizing, and organizing coding-agent trajectories, including benign and safety-critical scenarios. Support the implementation of agent-safety RL training for SWE-agent-like coding agents using base models from the Qwen3.5 family. Assist in building and maintaining controlled sandbox environments for trajectory sampling, training, and evaluation. Support the evaluation of safety improvement and utility preservation, including running experiments and summarizing results. Support manuscript preparation and assist in proposal and grant-related documentation. Requirements: A Bachelor's degree in Computer Science, Artificial Intelligence, Cyber Security, or a related discipline. Research or applied experience in reinforcement learning, LLM training, AI safety, or coding agents through coursework, projects, internships, or research activities. Basic experience with or strong interest in LLM-based coding agents, agent harnesses, and tool/terminal execution. Knowledge of RL training pipelines, dataset construction, or sandboxed execution environments; familiarity with PyTorch is an advantage. Proficiency in Python, with familiarity with shell/Linux environments. Ability to support academic writing and research outputs for international conferences. We regret to inform that only shortlisted candidates will be notified.