야, 너두 GPU에 LLM 배포할 수 있어: 온프레미스 GPU 기반 k8s 환경의 LLMOps 최적화 여정 | Model & Agent
kakao tech · 25:00 · Watch on YouTube
Kakao Pay runs open-weight LLMs like GLM-5.2 on on-premise GPU Kubernetes with vLLM because on-premise becomes cheaper than cloud/API after ~1 year and at scale, then optimized cold starts, pooling, KV-cache, and rout...