10. AI Platform Engineer
Evidence label: Role family. AI Platform Engineer now appears as an exact title across finance, media, healthcare, consulting, and software employers.
What they do
Build and operate the shared services that product teams use to create, evaluate, deploy, govern, and observe AI applications. Current postings make the AI scope explicit rather than treating it as a minor extension of generic platform engineering.
BlackRock's current description covers shared APIs, SDKs, agent runtimes, evaluation systems, observability, retrieval, guardrails, and governance. Realtor.com, Absa, and Alcon show the same ownership boundary with different cloud, cost, and regulatory constraints.
Responsibilities
- Provide standardized LLM integration patterns (internal SDKs, proxies, abstractions)
- Manage API keys, rate limits, and cost allocation across teams
- Build AI observability infrastructure (tracing, logging, alerting)
- Enforce security policies for AI outputs (PII filtering, output validation)
- Maintain model registries and versioning systems
- Create "paved roads" for RAG patterns, agent architectures, eval pipelines
Organizational boundary
The platform team owns reusable infrastructure, developer experience, and enforceable controls. Product or domain teams own each use case, acceptance criteria, and operational outcome. Centralizing use-case decisions in the platform team creates a queue; decentralizing credentials, observability, and guardrails creates inconsistent controls. The boundary should separate shared mechanisms from product accountability.
Required skills (AI additions)
MLOps tooling, LLM gateway products (LiteLLM, Portkey), cloud AI services, cost optimization patterns, security for AI (prompt injection mitigation, output filtering).