QYTANGQYTANG ACADEMY
中/EN
NVIDIA AI Factory Full-Stack Practice: Building an Enterprise LLM Production Line (NCP-AIO)

QYTANG VIP PROGRAM · Nvidia

NVIDIA AI Factory Full-Stack Practice: Building an Enterprise LLM Production Line (NCP-AIO)

Public preview1 videos14 curriculum items
COURSE INTRODUCTION

Course Introduction

This course builds an enterprise AI infrastructure and large-model production line from the ground up, covering networking, containers, Kubernetes, GPU clusters, image distribution, data storage, model training, evaluation, distributed inference, performance optimization, and observability. It combines NVIDIA official solutions with open-source approaches to develop reproducible full-stack AI implementation skills.

CURRICULUM

Course Curriculum

14
  1. 01

    Accounts, DNS, NTP, CA, .env, and Domain Names

  2. 02

    BCM HeadNode GUI Installation and Basic Tools

  3. 03

    cnode Image, Harbor Trust, NVIDIA User Space, and Pyxis/Enroot

  4. 04

    DOCA-OFED, PXE Provisioning, and Hardware Acceptance

  5. 05

    Docker, Kubernetes, GPU/Network/NIM Operators, BCM Gateway, and MIG

  6. 06

    Harbor Installation, Projects, Image Preloading, and Acceptance

  7. 07

    MinIO Installation, Buckets, and Object Paths

  8. 08

    Lustre Server/Client and Kubernetes PV/PVC

  9. 09

    GPU/MIG Labels, Run:AI Control Plane/Cluster, Projects, and Lab Spaces

  10. 10

    NeMo Platform, Asset Import, NCCL, Customizer, MIG Evaluator, and NIM

  11. 11

    Dynamo TensorRT-LLM Production Path, Four KV Scenarios, and Ideal TP Rank/GPU/NIC/Rail Mapping

  12. 12

    BCM Prometheus/Grafana Monitoring for Live Dynamo

  13. 13

    Slurm Fundamentals, CPU Examples, Pyxis/Enroot, and Dual-RTX Training

  14. 14

    SGLang PD Disaggregation, Four KV-aware Routing Scenarios, OpenAI API, and Grafana Dashboard

COURSE RECORDINGS

Course Video

Watch the public preview here or open it on the original video platform.