Uday maps AI‑infra: GPU pools, Pinecone

- Uday recommended an AI infrastructure stack of shared GPU pools, vector DBs like Pinecone, and Kubernetes inference for predictable developer workflows. - He argues GPU pooling reduces idle spend, Pinecone eases embedding retrieval, and K8s inference simplifies rollouts across teams and regions for scale. - The blueprint prioritizes self‑service infra that matches developer expectations for latency and cost in production. (x.com)