How Xoople Scales Python for AI using Anyscale on Azure | LIVE148

Milos Colic shares how Xoople scaled Python-based AI workloads on Azure using Ray via Anyscale, covering the distributed-systems challenges behind data ingestion, training, and inference, and why the team prioritized delivering outcomes over operating clusters.

Overview

This Microsoft Build 2026 customer conversation focuses on scaling production AI workloads on Azure with a managed Ray platform (Anyscale).

AnyScale as a managed Ray platform on Azure

• The session frames Anyscale as a managed way to run Ray on Azure.
• The core theme is reducing the operational burden of running distributed clusters so engineering teams can focus on product and AI outcomes.

Xoople’s AI stack on Azure

• The discussion covers how Xoople’s AI stack spans:
  • Data ingestion
  • Model training
  • Model inference
• It describes the evolution from early distributed Python approaches to running production workloads with Ray on Azure.

Hybrid and heterogeneous computing

• The speakers highlight benefits of hybrid and heterogeneous computing in Xoople’s AI operations.

Scaling large-scale image processing

• The session calls out scaling image processing workloads across extremely large datasets (described as trillions of pixels), using distributed Python systems.

Team collaboration and AI tooling

• The conversation touches on cross-team collaboration between product, application, and platform teams.
• It also mentions adoption of AI tools such as Copilot on Azure across teams (without going into product-specific setup details).