Organizing an AI Platform team: who owns what?

Discussion on team topologies: - Platform: Model routing, evals, observability, cost management, guardrails. - Product pods: Prompts, workflows, domain data, UX. - Data: Labeling, curation, and data contracts. - Security: Privacy reviews, red-team, incident response. Questions: Where do you place PMs and applied scientists? How do you avoid platform becoming a bottleneck while keeping standards high?