AI工程师（进攻性安全）
AI Engineer (Offensive Security)
在CovertSwarm，我们正在做很少有团队能实现的事情：为真正的伦理黑客开发代理工作流。这是一项从零开始的工程工作，你将与经验丰富的安全顾问合作，将他们的判断转化为由LLM驱动的系统，这些系统可靠、可观察，并经过严格的评估。
你将加入一个小型的新团队，不仅帮助我们构建产品，还参与塑造我们的工作方式。这一切都建立在低自尊、高信任和持续学习的文化之上。这是对真正自主和技术挑战充满热情的人的理想职位。
这是一个远程职位，面向位于欧盟或东部标准时间（EST）时区的候选人。我们每季度会聚一次，进行偶尔的线下会议，启动新的工作流并统一团队目标。
我们正在寻找3名新的AI工程师加入我们不断壮大的国际团队。
你将参与的工作
你将帮助设计、构建和改进支持真实进攻性安全工作的AI系统。这可能包括：
- 自主的黑客代理，能够端到端地执行多步骤的进攻性工作流程。你将设计和构建使它们可靠且可重用的架构。
- 将这些代理呈现给顾问的后端和界面：API、服务和UI，将原始代理功能转化为顾问可以独立使用的工具。
- 评估和基准测试，构建真实的靶场实验室和评估框架来衡量代理性能，以及验证逻辑来确认漏洞并减少误报。所有内容都基于可测试、可观察和可维护的系统。
- 为顾问提供的生成式AI功能，例如聊天机器人界面、文本生成工具和策略构思功能，解决实际交付问题。
- 应用型AI研究，区分真正有用的内容和炒作。你将定期有研究时间，并有无限的培训预算用于持续学习
技术栈
我们当前的技术栈包括Python、TypeScript、LangGraph、Docker和AWS。你不需要对所有这些技术都有深入经验，但你应该能够快速构建高质量的生产系统。
大部分工作位于工程、代理AI、评估、可观测性和可用的内部工具之间的交叉点。
你是什么样的人
你是一位在LLM和代理方面有深入经验的优秀工程师。你编写高质量的代码，思考严谨，能够从问题出发，全程推进解决方案。
你需要具备的经验才能成功
1. 强大的生产工程经验
查看英文原文
At CovertSwarm, we're doing something few teams get to do: building agentic workflows for real ethical hacking. This is greenfield engineering where you'll work alongside expert security consultants, translate their judgement into LLM-powered systems that are reliable, observable, and backed by rigorous evals.
You'll join a small, new team and help shape not just what we build but how we build it. All this, in a culture of low ego, high trust and constant learning. It's the perfect role for someone excited by real autonomy and technical challenge.
This is a remote role open to candidates based in EU or Eastern Standard Time (EST) time zones. We gather in person around once per quarter for occasional offsites to kick off new workstreams and align as a team.
We are looking for 3 new AI Engineers to join our growing international team.
What you'll work on
You'll help design, build, and improve AI systems that support real offensive security work. This may include:
- Autonomous hacking agents that carry out multi-step offensive workflows end to end. You'll design and build the architecture that makes them reliable and reusable.
- Backend and interfaces that surface these agents to consultants: APIs, services, and UIs that turn raw agent capability into tools consultants can use independently.
- Evaluation and benchmarking, building realistic target labs and eval harnesses to measure agent performance, and the validation logic that confirms exploits and reduces false positives. All of it built on testable, observable, and maintainable systems.
- Gen-AI features for consultants, such as chatbot interfaces, text-generation tools, and strategy ideation features that solve real delivery problems.
- Applied AI research, separating what's genuinely useful from what's hype. You'll have regular research time and an unlimited training budget for continuous learning
The tech stack
Our current stack includes Python, TypeScript, LangGraph, Docker and AWS. You don’t need deep experience in all of these, but you should be comfortable building production-quality systems and learning quickly.
Most of the work sits at the intersection of engineering, agentic AI, evaluation, observability, and usable internal tooling.
Who you are
You're a strong engineer who's gone deep on LLMs and agents. You write good quality code, you think rigorously, you take problems and run with them end to end.
What experience you need to be successful
1. Strong production engineering experience
You should have proven experience writing high-quality production code and building reliable applications across at least a portion of our tech stack. This includes writing testable and maintainable software, designing systems end to end, working with CI/CD, deploying services, debugging production issues, and making sensible engineering trade-offs.
2. Experience building and evaluating agentic workflows
You must have hands-on experience building agentic workflows or LLM-based systems that involve tool use, multi-step reasoning, orchestration, evaluation, or automation.
We are especially interested in people who have thought deeply about how agents fail, how to measure performance, and how to move beyond impressive demos into reliable systems. This experience may come from professional work, open-source contributions, research, or serious side projects.
3. Ownership, judgement, and clear communication.
You’re excited to work on a frontier problem where AI is being applied to real offensive security workflows in new and meaningful ways. You can take a rough idea, shape the approach, make good technical decisions, build and ship the system, and iterate based on feedback. You keep the wider goal in view, test assumptions quickly, adapt as we learn, and communicate clearly throughout, whether you’re documenting an architecture decision or explaining an agent failure mode to a non-AI specialist.
The Perks
Join a team that values both excellence and balance:
- True remote flexibility - work from anywhere.
- Unlimited training to keep your skills sharp.
- Unlimited vacation - because burnout helps no one.
- Private medical insurance and pension scheme.
- Conference speaking bonuses.
- A culture of radical candor, continuous improvement and technical excellence.
The Culture
At CovertSwarm, we take pride in pushing the boundaries of offensive security. Our team consists of passionate and humble professionals who value creativity, technical depth and delivering results that matter.
Ready to join the Swarm?
Take the next step in your career by applying today. Let’s talk about how your skills, research mindset and offensive capability align with CovertSwarm’s mission to redefine offensive security.