Staff Product Manager, Agentic AI ApplicationsActive$172K

Location: San Francisco or Mountain View, California

What you'll do

• Own the Agentic Platform strategy and roadmap. Define what ships, in what: order, and why. Translate organizational outcomes into concrete platform capabilities with measurable success criteria.

• Define and drive the agent and runtime. Establish the managed agent runtime: supporting multi step orchestration with durable execution, model gateway abstraction across all providers, governed tool invocation, and configurable per-agent guardrails (cost ceilings, timeouts, blast radius limits).

• Build the MCP connector ecosystem. Own the strategy for standardized,: bidirectional connectors various systems of record. Drive on behalf of identity propagation, idempotency, dry-run/preview mode, and a connector SDK that lets domain teams onboard new systems without platform changes.

• Establish the intelligence layer. Define the three layer data architecture:: knowledge graph (curated domain knowledge), context graph (live entity state from systems of record), and temporal memory (session, user preferences, and episodic history). Ensure unified retrieval across vector, structured, and graph sources with source traceability on every context element.

• Ship the evaluation and quality framework. Own the AI-judge evaluation: pipeline: offline eval with golden datasets, online LLM-as-judge scoring, domain-specific judges (Finance, HR, Legal, Sales), and mandatory evaluation gates in CI/CD. No agent reaches production without passing quality and safety thresholds.

• Design the developer experience. Make the platform self-service by: construction. Domain teams provision agent projects, promote across environments, and access connectors without platform-team tickets. SDK, CLI, sandbox environments, agent templates, and documentation — the paved road must be faster than building bespoke. Target: idea to production in Define the federation and adoption model. Establish the three tiers of adoption (platform built, domain built on platform, citizen developer edge apps) with governance checkpoints at each gate. Drive the gold-standard promotion pipeline from edge prototype to production hardened service.

What they're looking for

• + years of product management experience, with at least 3 years on internal: platform, infrastructure, or developer-experience products.
• Deep experience building platforms that other teams build on you understand: the difference between a platform and an application, and you have opinions about API design, developer ergonomics, and self service.
• Demonstrated experience with AI/ML platforms, agent frameworks, LLM-powered: applications, or agentic systems. You know what an agent runtime is, what RAG means in practice, and why evaluation is the hardest part.
• Strong technical foundation you can read architecture diagrams, discuss trade: offs with engineers (e.g., sync vs. async, checkpointing strategies, context window management), and make informed prioritization decisions on deeply technical work.