Cybersecurity · New York, NY

On-site Full-time

Security Architect, AI Infrastructure

Fluidstack · New York, NY · Posted Oct 21, 2026

This posting comes from Fluidstack. Confirm the details on their site before applying.

Overview

You're securing the frontier of AI. The model weights training on our infrastructure are the most valuable and most targeted artifacts in technology, and we're standing up the compute to hold them faster than anyone ever has. A breach isn't a leak, it's the frontier walking out the door. Build the entire security program from scratch. Most leaders inherit someone else's system and spend a career patching it.

Responsibilities

  • Own the attack surface unique to frontier compute: model weights custody, training environment isolation, accelerator and interconnect security, and multi-tenant boundaries under real adversarial pressure.
  • Design the reference architectures and standards that security engineering builds to, keeping the seam between design and build clean enough that both sides move fast.
  • Treat the workload as part of the threat model, and architect environments that stay defensible even when the thing running inside them is the adversary.
  • Run security design review for the systems that touch training and inference environments, from bare metal and fabric to scheduler and storage.
  • Track where frontier compute is going (new accelerators, interconnect generations, confidential computing, agentic workloads) and have the architecture ready before it lands.

Requirements

  • You've secured AI, ML, or HPC infrastructure where the assets were genuinely worth stealing: lab infrastructure security, supercomputing centers, or research environments that drew nation-state attention.
  • You can architect isolation at every layer it exists: hardware, firmware, hypervisor, kernel, container, network fabric, and physical.
  • You understand AI compute infrastructure as a system (accelerators, high-speed interconnects, parallel storage, schedulers) and know where each component breaks a classical security assumption.
  • You've done original work in systems or ML security, published or shipped, and can turn research into architecture that engineers actually build.
  • You defend design decisions in review on their merits, and change them when someone closer to the metal proves you wrong.
  • Bonus: Confidential computing and attestation depth. Weights custody or cryptographic key custody at scale. A publication record in systems or ML security.

Benefits

  • Bonus

Great fits start here.

Whether you're looking for your next role or your next team member, give people and possibilities a place to meet.