Security Engineer

present of AI; San Francisco (CA); On-site; USD 150,000 - 210,000; Full time

Job summary

Cognition in San Francisco is seeking a hands-on Senior Security Engineer who writes real code in Python, Rust, or Go. You’ll lock down sandboxes running untrusted code, review designs for weaknesses, and secure cloud and Kubernetes environments while leading incident response.

The ideal candidate works across product, infrastructure, and detection, stays calm during incidents, and helps shape threat models.

Qualifications

• Hands-on security engineering with ability to write production code.
• Experience with Python, Rust, or Go.
• Strong understanding of sandboxing, isolation, and threat modeling.
• Experience securing Kubernetes and cloud environments.

Responsibilities

• Lock down sandboxes that run untrusted code.
• Review designs for weaknesses and harden systems.
• Secure cloud and Kubernetes setups.
• Lead incident response when security events occur.

Skills

Python

Rust

Go

Security engineering

Tools

Kubernetes

Job description

Engineering San Francisco FullTime Senior level

This role asks for a hands-on security engineer who also writes real code in languages like Python, Rust, or Go. The work covers locking down sandboxes that run untrusted code, reviewing designs for weak spots, securing cloud and Kubernetes setups, and leading incident response. A strong candidate has worked across product, infrastructure, and detection, and is calm when something breaks.

How it fits the AI era

Devin runs code on behalf of users, so the main security question is how to let an AI agent act freely without letting it cause harm. AI creates new risks that few teams have solved, such as agents using tools and credentials over long tasks. Human judgment stays central for threat modeling, deciding which risks are acceptable, and leading a response when something goes wrong.

How it builds AI experience

The work puts someone right at the edge of AI agent security, where sandboxing, isolation, and tool access for autonomous coding agents are still being worked out. Candidates can prepare by studying prompt injection and agent isolation, building a small sandbox for running untrusted code, or securing a multi-tenant Kubernetes setup. Practice with detection and incident response on systems that run user code would also carry over well.

Worth showing

• Sandboxing and isolation design
• Kubernetes and cloud security
• Threat modeling and design reviews
• Incident response leadership
• Python, Go, or Rust projects

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Cognition

Applied AI lab building Devin, an autonomous software engineering agent that plans, writes, debugs, and deploys code