Staff Software Engineer, Ads Agent Infra

Pinterest, San Francisco, CA +2

What you will do

Architect, build and maintain durable execution platforms supporting crash-recoverable, stateful AI agent workflows.
Design standardized Model Context Protocol (MCP) servers, SDKs and agent runtime interfaces for cross-functional engineering adoption.
Establish LLM observability, execution tracing, prompt versioning and token-based cost governance across multi-agent systems.
Drive technical strategy and write foundational design docs that align infrastructure, security and product teams.
Use AI to accelerate design prototyping, documentation and code iteration while applying judgment and verification to ensure correctness and quality.
Automate repeatable testing, evaluation and grader pipelines for agent skills and memory services.

Pinterest, San Francisco, CA +2

$208K–$365K

a year

This base pay is above the median for staff infrastructure, platform and SRE jobs that are remote in the US.

We compare the middle of each posted range. This is base pay as posted. The group does not include this job.

The Pinterest site opens in a new tab.

See total pay at Pinterest on Levels.fyiBase, stock and bonus, as employees report it.

No one has reported an interview at Pinterest yet. Did you interview there? Add the questions.

Compare your résumé with this job. See where you match, what is missing and which lines to change.

Architect, build and maintain durable execution platforms supporting crash-recoverable, stateful AI agent workflows.
Design standardized Model Context Protocol (MCP) servers, SDKs and agent runtime interfaces for cross-functional engineering adoption.
Establish LLM observability, execution tracing, prompt versioning and token-based cost governance across multi-agent systems.
Drive technical strategy and write foundational design docs that align infrastructure, security and product teams.
Use AI to accelerate design prototyping, documentation and code iteration while applying judgment and verification to ensure correctness and quality.
Automate repeatable testing, evaluation and grader pipelines for agent skills and memory services.