Yi Wang, Tech Lead of AI Platform, Baidu, NYC 2017

on YouTube ↗22m

" summary

Integration of Baidu’s PaddlePaddle deep learning framework with Kubernetes to run large-scale, fault-tolerant training on general-purpose clusters. Covers how Kubernetes is used instead of MPI-based cluster management so deep learning can coexist with web services (Nginx), log collectors (fluentd, Kafka), and data processors (Hadoop, Spark, Storm) in an industrial data pipeline.

AI-written summary · about

" chapters (6)

1. 0:00Start
2. 5:48CLOUD AND ON-PREMISE CLUSTERS
3. 7:00THE SOLUTION - GENERAL PURPOSE CLUSTERS
4. 9:38CHALLENGES - GENERAL PURPOSE CLUSTERS
5. 11:46CHALLENGES - FAULT-TOLERABLE JOBS
6. 16:06AUTO FAULT-RECOVERY

" related talks

• 39mEvolving Deep Learning Platform with KnativeTi Zhou · KubeCon + CloudNativeCon China 2019
• 39mEvolving Deep Learning Platform with KnativeTi Zhou · KubeCon + CloudNativeCon + Open Source Summit China 2019
• 18mKeynote: Optimizing AI Workload Scheduling: Bilibili's Journey To an Efficient Cloud Native AI PlatformKevin Wang, Long Xu · KubeCon + CloudNativeCon China 2025
• 21mKeynote: A Tale of Using Kubeflow to Make the Electricity Smarter in ChinaXin Zhang, Julia Han · KubeCon + CloudNativeCon China 2018
• 09mKeynote: Kubernetes for GenAI: The New Operating System of the AI-First WorldJanakiram MSV · KubeCon + CloudNativeCon India 2025
• 15mKeynote: From Platforms to AI Factories. Has Kubernetes Solved It?Saiyam Pathak · KubeCon + CloudNativeCon India 2026
• 06mSponsored Keynote: Building Your AI Factory: Wh...Andreea Munteanu · Cloud Native AI + Kubeflow Day 2026
• 31mThe Future of Data on Kubernetes: From Database Management To AI FoundationGabriele Bartolini, Rob Strechay, Nimisha Mehta, Brian Kaufman · KubeCon + CloudNativeCon Europe 2025