[ research ]
Research Engineer, Safety & Alignment
Cognition- Location
- Not specified
- Work mode
- On-site
- Type
- full time
- Listed
Applications go directly to Cognition. Ramai takes no fee for placement and no money for ranking.
[ About this role ]
Research Engineer, Safety & Alignment at Cognition. This role involves building agentic safety evaluations, red-teaming the Devin AI agent, and developing alignment methods for model training. The focus is on hands-on research engineering for safety work integrated into a real agent.
Skills and focus areas
- safety evaluations
- red-teaming
- alignment methods
- reward modeling
- preference data
- constitutional approaches
- agent harness
- prompt injection
- data exfiltration
- sandbox escape
- reward hacking
- misuse
Summary written by Ramai from the original posting. Full details and requirements are on Cognition’s careers page.
[ More at Cognition ]
See all 106 open roles at Cognition →[ Similar roles at other startups ]
Browse all open research roles →[ Not quite right? ]
Tell the assistant what you’re after — stage, location, stack, visa sponsorship — and it searches every open role for you. A few questions are free, no account needed.