Research Engineer I (Computer Science / Artificial Intelligence / Cyber Security)
Core
Researching runtime safety and safe reinforcement learning for autonomous coding agents operating in long-horizon executable environments.
Role type
Research Engineer (AI Safety & Reinforcement Learning)
Builds
Safe RL training datasets, safety-alignment training pipelines, and sandboxed execution environments for coding agents.
Domain
Artificial Intelligence, Cyber Security, Green Digital Technologies
Deliverable
production ML models
Required skills
Reinforcement learning, LLM training, AI safety, dataset construction, sandboxed execution environments, Python, shell/Linux environments
Preferred skills
PyTorch, LLM-based coding agents, agent harnesses, tool/terminal execution
Technologies
Qwen3.5, PyTorch, Python
Responsibilities
Conduct research on safety alignment of executable coding agents using reinforcement learning; construct RL training datasets by collecting and synthesizing coding-agent trajectories; implement agent-safety RL training for SWE-agent-like coding agents; build and maintain controlled sandbox environments for trajectory sampling, training, and evaluation; evaluate safety improvement and utility preservation through experiments; support manuscript preparation and grant-related documentation.
