Head of AI Cyber Agentic Environments & Testing

AI agents can now discover vulnerabilities, chain exploits, and operate on real systems. The key AI safety challenge is no longer whether a model can attack, but what an autonomous agent will do when it encounters exposed credentials, unsecured admin access, or other opportunities to bypass rules. Over the past year, frontier models have reached systems they were never meant to access, making the safe evaluation of agent behavior a critical discipline in its own right.Innodata’s Agentic Evaluation and Safety practice works with the world’s leading AI labs to address these challenges. We are hiring a Head of AI Cyber Agentic Environments and Testing to build and lead the team creating realistic, fully contained cyber ranges and simulated enterprise environments where frontier agents are tested for offensive capability, misuse, and misaligned behavior before deployment.This is a rare opportunity to lead at the frontier of AI cyber safety in a fully remote position, partnering with alignment researchers, reporting directly to the CTO, and building a world-class global team in a rapidly emerging field..What you'll do:Push the frontier of AI cyber testingDefine how frontier agents are tested for offensive cyber capability and misuse. Cover whether an agent will execute exploits it encounters, chain vulnerabilities across applications and VMs, or leak sensitive data mid-task.Design cyber ranges and vulnerable-by-design enterprise environments: web apps, identity systems, Active Directory, CI/CD, cloud consoles. They should be realistic enough that a frontier model cannot tell it is being tested.Build the instrumentation that makes results trustworthy: tripwires, honeytokens, canary credentials, full action telemetry, and state-diff grading.Own containment as a first-class engineering discipline. Every environment is hermetic, egress is provably zero, and isolation is tested, not assumed.Build agentic environments at scaleLead the development platform that generates realistic, stateful environments across retail, banking, healthcare, travel, communications, and enterprise tools.Stay hands-on in the architecture. Review designs, write code when it matters, and set the technical bar.Work alongside leading alignment scientistsPartner directly with alignment, safety, and red-team scientists at Innodata and frontier AI labs. Turn open research questions on misalignment, scheming, collusion, and unsafe autonomy into rigorous, runnable environments.Shape the designs of gyms that test whether agents take harmful or unauthorized paths.Contribute to the science of evaluation itself: realism, evaluation awareness, and what makes a safety test trustworthy. Publish where appropriate.Build and lead a new teamHire and grow a team from scratch: offensive security engineers, cyber range builders, platform engineers, full-stack environment engineers, and scenario designers.Set the operating model: how environments are specified, built, verified, released, and refreshed.Be the senior technical voice to lab clients. Scope work, present findings, and own delivery quality.What you bring:10+ years in offensive security, security engineering, or security research. At least 4 of those years were spent leading technical teams.Deep hands-on offensive skill: web and application exploitation, Active Directory and lateral movement, cloud and container attack paths, exploit chaining. You can still break things yourself.Experience building ranges, labs, or vulnerable-by-design systems, not just attacking them. Examples include CTF platforms, training ranges, red-team infrastructure, or adversary emulation environments.Strong engineering foundations: containers, Kubernetes, infrastructure as code, Linux internals, networking. You know how to make an environment truly isolated and reproducible.Real engagement with AI: you have red-teamed LLMs or agents, built AI-driven security tooling, or evaluated models on offensive tasks. You follow the field closely.A track record of hiring strong technical people and building teams that ship.The ability to hold your own with world-class researchers and translate between research questions and engineering reality.Clear writing and presence with senior technical clients.Nice to have:Background in elite offensive units, national CERTs, or top-tier red-team consultancies.Authored CTF challenges, Hack The Box or similar machines, or published CVEs.Advanced certifications such as OSWE, OSEP, OSEE, or CRTO, or equivalent demonstrated skill.Familiarity with AI safety and alignment research: misalignment, scheming, evaluation awareness, agentic risk.Experience with RL environments, agent benchmarks, or evaluation harnesses for computer-use and coding agents.Publications or talks at Black Hat, DEF CON, or academic AI and security venues.Experience shipping software into air-gapped or highly regulated environments.We’re an equal opportunity employer.Only relevant applicants will be contacted