Log inSign up
DevopsCube profile banner
user avatar

DevopsCube

@devopscube
DevOps, MLOps, and Cloud tips & Deep Dives
Join 20,000+ Readers →
newsletter.devopscube.com
Joined November 2014
10
Following
11.3K
Followers
RepliesArticlesMedia
  • Pinned
    user avatar
    DevopsCube
    @devopscube
    Aug 11
    Many DevOps concepts map directly to MLOps. Here is one example from MLflow. Data scientitst/ML engineers compare multiple training runs and select the best-performing model. Then you can promote it for deployment by registering it in the MLflow model registry. During
    Image
    GIF
  • user avatar
    DevopsCube
    @devopscube
    28m
    The MLOPS for DevOps repo is close to 500 stars 🚀 We started this series with a simple goal. To help DevOps engineers understand the key concepts and workflows required for MLOps. Because knowing just Terraform, cloud, and Kubernetes is not enough for MLOps. You need
    Image
  • user avatar
    DevopsCube
    @devopscube
    Aug 12
    How does a trained AI model become a production API on Kubernetes? KServe is a CNCF incubating project for serving AI models on Kubernetes. Simply put, KServe takes a trained model and turns it into a scalable inference service on Kubernetes. It handles deployment,
    Image
    GIF
  • user avatar
    DevopsCube
    @devopscube
    Aug 10
    Canary Vs Shadow Deployment vs. A/B Testing When we deploy a model to production, It also follows the deployment strategies we usually use for regular applications. Let's understand the three common model deployment strategies. 𝟭. 𝗖𝗮𝗻𝗮𝗿𝘆 𝗧𝗿𝗮𝗳𝗳𝗶𝗰
    Image
  • user avatar
    DevopsCube
    @devopscube
    Aug 9
    🛠 DevOps tool of the Week: Ray ML workload works fine on one GPU. The complexity comes when you want to scale it across GPUs running on multiple nodes. That's where Ray helps. Ray is an open-source framework with 43K+ stars on Github. It helps Ray helps handle distributed
    Image