Software Engineer, AI Product Security

Software engineer who sets the security foundations for Notion's AI products, partnering with the teams that build them on how agents run, which tools they may call, how retrieval and workspace edits are controlled, and how the whole surface resists abuse. The engineer turns threat models into architecture, launch criteria and production checks, ships reusable libraries, test fixtures and guardrails, builds automated red-team and regression tests for prompt injection, exfiltration, tool misuse and cross-tenant leaks, and helps set the bar for safe internal use of coding agents and MCP tools.

Notion wants eight to ten or more years designing, building or securing complex software, strong judgement about trust boundaries and data flow, and a builder's habit of leaving tools behind. The role is hybrid in San Francisco, in the office on Mondays, Tuesdays and Thursdays.