The RoleEnterprises want to deploy AI agents that read internal data, call tools and take actions. Their security teams often lack the means to approve these agents with confidence. AgentWarden secures agents across that lifecycle, from risk assessment before deployment through runtime monitoring and incident response. As its Product Manager, you will own the problem of getting agents approved and operating safely in production. You will work with our engineering and research teams and directly with customers' security, platform and AI teams.
In your first months you will focus on agent risk assessment. This includes identifying risky combinations of tools and permissions, such as agents that pair access to private data with exposure to untrusted content and the ability to communicate externally. It also includes analyzing the third-party tools agents connect to and how agents behave when they run. You will spend significant time with customers as well: narrowing down which agent risks block their deployments, identifying economic buyers, and turning pilots into scoped product requirements.
What You'll Work On- Pre-deployment risk assessment for agents: analysis of agent tools, permissions and runtime behavior, and the evidence security teams need to approve an agent.
- Red teaming for agents, covering multi-step attacks, tool misuse and indirect prompt injection.
- Runtime controls for agent actions and tool calls, kept current as new attack techniques are published.
- Alerting and investigation for agent incidents, where a single session can span many steps and systems.
Responsibilities- Own the AgentWarden roadmap against a defined set of customer problems and success metrics.
- Lead customer discovery and pilot engagements with enterprise security and AI platform teams, and present to senior security stakeholders.
- For each deployment, identify the buyer, the approver and the operator, and shape pilots around what each of them needs in order to proceed.
- Write requirements that engineering and research can build and evaluate against, including test sets and acceptance criteria.
- Track agent frameworks, MCP, agent platforms and related threat research, and update the roadmap as they change.
- Work with the DynamoGuard and DynamoEval PMs on shared policies, evaluations and alerting.
Requirements- 2 to 4 years of product management experience, preferably in security, developer tools, or AI/ML products sold to enterprises.
- Hands-on familiarity with how agents are built, including tool calling, MCP, and at least one agent framework such as LangGraph, the OpenAI Agents SDK or CrewAI. You should be able to build a simple agent yourself.
- Working knowledge of LLM and agent threats, such as prompt injection, excessive agency and data exfiltration, and of references such as the OWASP Top 10 for LLM Applications.
- Ability to read Python or TypeScript well enough to review scanner findings and discuss detection approaches with engineers.
- Experience running customer discovery with technical buyers and converting it into scoped requirements.
- Clear written and verbal communication, including presenting to enterprise security leaders.
- Willingness to spend 4 to 6 weeks per year at customer sites, and more during large contract engagements.
Nice to Have- Background in security engineering, application security, red teaming or detection engineering.
- Experience with products that produce findings or alerts for SOC or AppSec teams.
- Experience deploying into financial services or other regulated industries.
Why Join Us- Define how enterprises secure AI agents while the category is still forming.
- Work directly with Fortune 500 security teams on production deployments.
- Own a product line end to end at a founder-led startup.
- Competitive compensation, equity and benefits.
