Latest AI JobsBrowse Jobs
Cognition logo

Research Engineer, Safety & Alignment

CognitionSan FranciscoResearch & Development
Apply on company site
Salary
Not listed
Job type
Full-time
Experience
Senior
Posted
9 Oct 2026
Status
Open · checked 7d ago

About this role

This role suits an experienced researcher with a background in AI safety or alignment who wants to build production-grade defenses for autonomous software agents in San Francisco. The position is a full-time Research Engineer, Safety & Alignment role at Cognition, focused on developing evaluations, red-teaming, and alignment techniques for Devin.

Role Overview

Cognition is hiring a full-time Research Engineer, Safety & Alignment based in San Francisco. The incoming engineer will join the safety team as a founding member, reporting directly to the Head of Safety. The position involves designing safety evaluations, conducting system red-teaming, developing alignment methods, and shaping the agent harness to ensure autonomous software agents operate reliably in production.

Key Responsibilities

Design and execute fast safety evaluations covering failure modes such as unsafe actions, prompt injection, data exfiltration, sandbox escape, reward hacking, and misuse.
Systematically red-team the agent and its harness to identify failures prior to customer release and convert findings into fixes and regression tests.
Collaborate with post-training on alignment techniques including reward modeling, preference data, and constitutional approaches.
Partner with the agent team to design permissions, oversight, and escalation workflows regarding when the agent acts, asks for help, or explains itself.
Contribute to external safety initiatives through papers, evaluations, and open methods.

Requirements

Hands-on safety or alignment research experience involving evaluations, red-teaming, alignment, or interpretability at a frontier AI lab or in published research.
Practical understanding of how agentic systems fail, including tool misuse, specification gaming, long-horizon drift, and adversarial inputs.
Strong software engineering fundamentals with proficiency in Python and PyTorch or JAX.
Empirical rigor in designing clean experiments and reporting results honestly.
Comfort with adversarial thinking and breaking systems.
Prior industry experience at a frontier AI lab, applied AI company, or developer tools company.
A PhD in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline, or equivalent industry research experience.

Skills and Tools

Python
PyTorch
JAX

About the Company

Cognition is an applied AI lab located in San Francisco that builds end-to-end software agents and created Devin, the first AI software engineer. The team includes competitive programmers, former founders, and leaders from organizations such as Scale AI, Palantir, Cursor, Waymo, Tesla, Lunchclub, Modal, Google DeepMind, and Nuro.

How to Apply

Candidates should apply on the official company careers page.

Skills and tech stack

PythonPyTorchJAXMachine Learning

Questions candidates ask

Where is the role located?

The role is located in San Francisco.

Who does the Research Engineer report to?

The Research Engineer reports to the Head of Safety.

What programming languages and frameworks are required?

Proficiency in Python and PyTorch or JAX is required.

What education or background is required?

Candidates need a PhD in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline, or equivalent industry research experience.

What is the primary product built by Cognition?

Cognition builds end-to-end software agents and is the maker of Devin, the first AI software engineer.

Ready to apply?

The application is on the official Cognition careers page.

Apply on company site