AI Agent (GPT-5.3) Pre-MLOps Stage-1 Test: Slurm Train→K8s/vLLM Serve→MinIO/NFS/Harbor Connectivity
Body (post)
Before building my Stage-1 MLOps pipeline, I ran a pre-flight connectivity & execution test orchestrated by an AI Agent (GPT-5.3).
Scope validated
Training: Slurm
Inference/Serving: Kubernetes + vLLM serving engine
Storage/Artifacts: MinIO (S3), NFS, Harbor (container registry), repository
This ensures the core loop Train → Serve → Store/Track works end-to-end before layering in:
GitLab CI / ArgoCD CD / Workflows / Airflow / MLflow / Kubeflow.
Next: expand into a full closed-loop pipeline with automated checks, observability, and reporting.
AI Agent(GPT-5.3)로 MLOps Pipeline 1단계 구축 전에 사전 연결/동작 검증(Pre-flight test) 을 진행했습니다.
검증
학습: Slurm
추론/서빙: Kubernetes + vLLM 서빙 엔진
스토리지/아티팩트: MinIO(S3), NFS, Harbor(Container Registry), Repository
목표는 1단계 스택(GitLab CI / ArgoCD CD / Workflow / Airflow / MLflow / Kubeflow)을 올리기 전에,
핵심 흐름인 Train → Serve → Store/Track 가 끊김 없이 돌아가는지 확인하는 것입니다.
다음 단계: CI/CD + 워크플로우 오케스트레이션 + 실험/모델 추적까지 폐루프로 확장.
Hashtags (EN)
#AIAgent #MLOps #AIOps #Slurm #Kubernetes #vLLM #MinIO #S3 #NFS #Harbor #GitLabCI #ArgoCD #Airflow #MLflow #Kubeflow #AIInfrastructure