Security Engineer

Cognition Cognition · Coding AI · San Francisco, CA · Research & Development

Security Engineer role focused on securing AI software agents (Devin, Windsurf) that execute arbitrary code, handle sensitive customer data, and operate within developer environments. Responsibilities include hardening sandboxing, owning product and infrastructure security, building security tooling, leading incident response, and driving customer trust. Requires strong software engineering fundamentals, cloud security expertise, and experience with threat modeling and incident response, particularly in the context of novel AI security challenges.

What you'd actually do

  1. Secure the agent execution surface: Design and harden the sandboxing, isolation, and runtime controls that let Devin safely execute untrusted code and use tools across long-horizon tasks.
  2. Own product and infrastructure security: Lead threat modeling, secure design reviews, and vulnerability management across Devin, Windsurf, and the underlying infrastructure they run on.
  3. Build security tooling that engineers actually use: Create internal systems for secrets management, identity and access, dependency security, and detection that integrate naturally into how the team ships.
  4. Lead incident response and detection: Build the detection pipeline, run incident response, and turn every event into systemic improvements.
  5. Drive customer trust: Partner with go-to-market and legal teams to support compliance and customer trust initiatives. Build the controls that customers expect from a tool deeply embedded in their engineering workflow.

Skills

Required

  • Python
  • Rust
  • Go
  • Kubernetes
  • AWS
  • GCP
  • Azure
  • Threat modeling
  • Incident response

Nice to have

  • Security engineering
  • Product security
  • Infrastructure security
  • Detection and response
  • Cloud security
  • Web security
  • Adversarial thinking
  • Autonomous agents
  • AI-native developer tools

What the JD emphasized

  • security surfaces in the industry
  • arbitrary code
  • highly sensitive customer code, credentials, and infrastructure access
  • define what security looks like for AI-native developer tools
  • ship fast without compromising on safety
  • hands-on, high-leverage security work
  • edge of what is being figured out for the first time
  • harden the sandboxing, isolation, and runtime controls
  • safely execute untrusted code
  • long-horizon tasks
  • vulnerability management
  • secrets management
  • identity and access
  • dependency security
  • detection
  • incident response
  • customer trust
  • Deep security engineering
  • product security
  • infrastructure security
  • detection and response
  • complex systems codebases
  • Cloud security expertise
  • multi-tenant compute environments
  • Web security expertise
  • complex, modern web applications
  • Threat modeling and adversarial thinking
  • look at a system and quickly identify how it breaks
  • think like an attacker and design like a defender
  • Incident response
  • leading incidents end to end
  • driving the fixes that follow
  • novel problem spaces
  • security challenges unique to autonomous agents and AI-native developer tools
  • frontier AI lab
  • applied AI company
  • developer tools company

Other signals

  • AI agents
  • AI-native developer tools
  • security for AI systems