« Mission Impossible - Help Tom Cruise leverage OpenShift AI to stop the crazy train before it’s too late ! »
Monday 8 July 14:00 - 17:00
Room 128

Message to Ethan Hunt : “The train is running mad at full speed and has no driver ! Your mission, should you choose to accept it, is to train and deploy an AI model at the edge to stop the train before it crashes. This message will self-destruct in five seconds. Four. three. Two. one.  tam tam tada tum tum tada tum tum tada tum tum tada tiduduuuuummmm tiduduuuuuuuuummm (mission impossible theme)”

This session can be delivered as a breakout session or as a lab.

Lab version :

A Lego City train (with a webcam) on a circuit is set up on a table at the center of the room. If you place the stop sign and the train stops, you win ! Ten teams, four persons per team and a train running mad at full speed. The ten teams compete to use OpenShift AI to train a model in order to recognize the traffic signs and stop the train before it’s too late.

There are the four personas in each team: data engineer, data scientist, operations and machine learning teams. They will combine their efforts to implement an AI platform that covers all stages of the MLOps life cycle.

  • The operations persona will be in charge of creating the AI platform in a disconnected environment leveraging Nvidia GPUs. They will also configure an edge device with microshift to support the model deployment.
  • The data engineer will collect, transform and make reliable data available using streaming.
  • The data scientist will build an object detection model and create a data science pipeline. They will deploy the model and monitor related metrics and drift.
  • Finally, the machine learning engineers will optimize inference and automate the model deployment up to the edge.

Breakout version :

A Lego City train (with a webcam) on a circuit is set up on stage. The presentation describes the architecture designed to make the train run on auto-pilot :

  • AI model (YOLO) deployed on a Nvidia Jetson Orin Nano
  • Lego train driven using Bluetooth Low Energy
  • Nvidia Jetson Orin Nano running Red Hat Device Edge with Microshift
  • Data acquisition, transformation, aggregation using AMQ, Kafka and Apache Camel
  • Data presentation using Quarkus
  • Data labelling, model tuning & training
  • Model deployment at the edge