AI Agent Security Research Engineer

OKX is a major cryptocurrency exchange operating OKX Wallet, serving millions in crypto trading and institutional market access.

Core Responsibilities:

1. AI-Driven Code Security Detection — Build a multi-agent collaborative code auditing system for vulnerability detection, malicious code identification, and data leakage prevention. Design agent role decomposition (Planners/Executors/Critics), tool chains, and state synchronization mechanisms. Integrate RAG and Chain-of-Thought / Reflection techniques while developing GitLab CI/CD, Tekton, and Jenkins plugins.
2. LLM Application Security Framework — Construct protections across input (prompt injection), output (compliance), and runtime layers. Develop automated alert classification workflows and human-machine intervention mechanisms aligned with OWASP standards for large language models.
3. Engineering & Platform Services — Design scalable, highly available agent architectures with concurrent task scheduling and fault tolerance; standardize API outputs and build closed-loop feedback systems.

Required Skills:

• 3+ years backend development (Python/Go/Java proficiency)
• Hands-on production LLM Agent deployment experience
• Demonstrated AI security knowledge (prompt injection, jailbreaking defenses)
• Docker/Kubernetes expertise with microservices architecture
• Production experience with at least one of: LangChain, LlamaIndex, AutoGen, CrewAI, or LangGraph
• LLM fine-tuning (SFT, LoRA), Llama 3, Qwen, DeepSeek
• SAST/SCA tools (CodeQL, Semgrep, SonarQube)

Locations: Hong Kong SAR and Singapore (APAC)