Description:

The foundation of any normal application starts from our very own laptop. However, in the case of AI applications, that might not be feasible since you need more resources that your system might not be able to support. Not everyone can just buy a system for that purpose. How should one go about creating an AI application, for example, a custom-trained chatbot? To solve this issue, we can leverage Ray: an open-source unified framework for scaling AI and Python applications.

In this talk, we cover end to end build of distributed AI application..

Outline of talk

• Introduction to Ray Framework (3 min)
• Overview of Demo Application (2 min)
• How Ray Helps (2 min)
  • Show end-to-end architecture when Ray in picture
• Demo
  • Setting up RayCluster (2min)
  • Preparing Data (2min)
  • Training Model (7min)
  • Serving Model (5min)
• Whats next? (2)
  • Dynamic Model Selection
  • Periodic Training
  • Model Composition
• Q/A (5 min)

Prerequisites:

• LLM Basics
• Python Basics
• DevOps Basic
• Distributed System Basics

Speaker Info:

Sudhanshu Prajapati is a Developer Advocate @ InfraCloud, With expertise in cloud-native application and distributed systems, he’s a goto source for hands on knowledge. Apart from his work, he contributes to open source projects like Aperture, Botkube. While main stream interest lies in core engineering he also explores giving talk and primer for folks to get started and share his knowledge. You can ask him questions around Data, Cloud, and AI.

Speaker Links:

• Previous Talks
  • FOSSDelhi Meetup 2023
  • K8s Bangalore 2023
  • CNCF Gurugram 2024
  -
• GitHub@sudhanshu456
• X@Sudhanshu_456