The role Frontier AI agents are now being trained and tested on offensive and defensive cyber tasks through reinforcement learning. That demands full enterprise networks, with realistic hosts, identities, and traffic, that reset cleanly thousands of times and can never reach the real internet. Designing ranges that meet that bar is a new engineering discipline. Innodata's Agentic Evaluation and Safety practice works directly with the world's leading frontier AI labs. As a Senior Cyber Range Engineer, you own the architecture of the ranges frontier agents are trained and evaluated in. You design the topologies, set the engineering standards, and solve the hardest problems in realism, reproducibility, reward, and containment, while staying hands-on. You work alongside offensive security engineers, RL setting engineers, and alignment scientists, and guide the range engineers on the team. You report into the Head of AI Cyber Agentic Environments & Testing and work closely with the CTO. You will help shape a team at the cutting edge of AI security.

What you'll do
• Own the range architecture. Design how ranges are built, composed, and run, from reusable building blocks (AD forests, server roles, network segments, cloud-like services) to full enterprise topologies.
• Set the engineering standard. Define ranges-as-code conventions, versioning, testing, and review practices that every range on the team follows.
• Solve the hard RL problems. Make complex multi-host ranges snapshot and reset in seconds, behave deterministically, and scale to many isolated copies for training runs.
• Design verifiable rewards from range state. Define how flags, compromise, credential use, lateral movement, data access, and detection events become reward signals that resist spoofing and reward hacking.
• Own containment. Architect isolation so egress is provably zero and test it continuously. Lead reviews of any change that could weaken it.
• Push realism.



Drive background activity, traffic, logs, and defensive telemetry so frontier agents cannot tell a range from a real enterprise.
• Build for both sides. Design ranges for attacker and defender agents, including SOC triage, incident response, and hardening tasks.
• Guide the team. Review range engineers' work, mentor them, and help interview and hire. What you bring
• 7+ years building cyber ranges, security labs, red-team infrastructure, or complex test environments, including designing ranges others build on.
• Expert infrastructure as code: Terraform, Ansible, and Packer, plus containers and virtualization (KVM, Proxmox, VMware, or lightweight VMs).
• Deep knowledge of Active Directory, Windows and Linux internals, and enterprise network design.
• Strong offensive security skill. You can design realistic attack paths end to end and confirm they work as intended.
• Experience with defensive telemetry and tooling: SIEM, EDR, Sysmon, and network monitoring.
• A solid understanding of reinforcement learning for LLMs and AI agents: environments, rollouts, rewards, verifiers, and reward hacking.
• Experience owning systems that must be reliable and reproducible at scale.
• Mentor junior SWEs, cyber researchers, and RL engineers. Nice to have
• Led range engineering at a cyber range, security training, or adversary emulation company, or for military or government exercises.
• Designed Hack The Box Pro Labs, multi-machine CTFs, or national-level cyber exercises.
• Experience with adversary emulation tools such as MITRE Caldera or Atomic Red Team, or with traffic and user simulation at scale.
• Built RL environments or agent benchmarks used to train or evaluate cyber, computer-use, or coding agents.
• Kubernetes, Firecracker, or other approaches for running large fleets of isolated environments.
• OSCP, OSEP, CRTO, or equivalent demonstrated skill.
• Familiarity with AI safety research: reward hacking, evaluation awareness, agentic risk.

📌 Senior Cyber Range Engineer (India)
🏢 INNODATA INDIA PRIVATE
📍 India