Staff AI Platform Engineer

Responsibilities

• Drive the business-side Bedrock runtime and gateway configuration - model access, cost attribution, content-filtering guardrails, model pinning, and quota management, behind an AWS/Okta-SSO'd gateway - co-owned with our DevX lead, who leads the shared-substrate architecture.
• Build and run the MCP fleet - ship, secure, and patch self-hosted and managed MCP servers for business apps and databases, and own the MCP registry.
• Partner with IT to graduate MCPs into stable, IT-managed assets.
• Make the secure path the easy path without administering identity or devices yourself: author the MDM config payload and the Okta-group-to-IAM mapping that IT pushes and enforces, and specify the desktop client setup and auth flows for non-technical employees.
• Build the AI and agent-triggered automations (approvals, spend alerts, agent-initiated workflows); general business automation stays with IT.
• Own the technical direction for the agent-hosting substrate other teams build on - the design docs and standards, not just the code. There is no existing pattern today; you set it, and other teams adopt it as their default.
• Define what "healthy" and "safe to ship" mean for every agent and MCP server on the platform - the cost, usage, and reliability telemetry (the Bedrock-to-Snowflake pipeline, identity-matched, with threshold alerting), plus a tiered-autonomy model: what stays read-only, what can act, what needs a human in the loop or a…
• Build the governance that standard requires. Route workflows by data sensitivity, enforce least privilege and auditability, and keep unproven models sandboxed away from production data.
• Drive alignment across IT, Security, and DevX on the platform's direction - get multiple teams building against a written standard without owning their roadmaps.

Languages