AI大模型安全算法工程师
Core
Aligning foundational large language models for cybersecurity to identify malicious intent and define task boundaries across dialogue and agent execution scenarios.
Role type
Senior IC AI safety engineer (large language models)
Builds
Cybersecurity-aligned LLMs and evaluation frameworks
Domain
Artificial Intelligence + Cybersecurity
Deliverable
production ML models
Required skills
Transformer architecture, RLHF/DPO, Python, PyTorch, adversarial search, data distribution design, cybersecurity risk assessment, quantifiable safety metrics
Preferred skills
Red teaming, threat intelligence analysis, preference optimization, reasoning chain safety constraints
Technologies
PyTorch, LLMs, Agent frameworks
Responsibilities
Collaborate with base model teams to align cybersecurity safety; construct layered harmful/borderline/benign datasets using real risk cases and red-blue对抗; generate and mutate adversarial prompts via LLMs; build evaluation systems for refusal rates and false positives; research intrinsic safety vs external guardrails; track and engineer frontier safety research.
Seniority
Senior, hands-on IC