AI/ML Platform Engineer III

•    Build and operate the shared agent runtime: planning loops, memory, tool calling and agent-to-agent coordination.

•    Stand up and operate the MCP/tool registry, AI Gateway — a governed catalog of MCP servers and tools with scoped permissions — and A2A coordination surfaces.

•    Implement workflow and agent-step orchestration with retries, timeouts and human-in-the-loop checkpoints.

•    Build retrieval-augmented generation patterns for agent context: embeddings, vector indexing, semantic search and retrieval quality evaluation.

•    Deploy and operate open-source model serving frameworks (vLLM, SGLang) and NVIDIA NIM microservices on the platform's GPU environments, following patterns established.

•    Run day-to-day GPU serving operations: deployments, rollbacks, scaling and utilization/cost monitoring, escalating tuning and capacity decisions to senior engineers.

•    Provision sandboxed agent execution environments with DevOps and security engineering.

•    Wire guardrails enforcement and observability — traces, token usage, outcome logging — into agent flows by default.

•    Build evaluation and regression testing for agentic behaviors.

•    Contribute to the agent build-and-deploy golden path: templates, SDK examples and documentation.