ML in machinery is our main topic this week as Mark Mirchandani and Brian Dorsey talk with Arnaud Hubaux, development lead for ASML. Our good friend, fellow Googler, and ML specialist Dale Markowitz joins the conversation as well!

With a potent combination of physics and Machine Learning, ASML develops machines that build the chips powering our essential devices. These highly refined machines optimize production for each specific customer, detect defects, and make corrections quickly and accurately by harnessing the power of Machine Learning. The machines also effectively train themselves to ensure the intense accuracy required is sustained over time. Arnaud describes the process these machines go through in order to accomplish these goals, including how they build and train their ML models with the help of physics.

Arnaud tells us how the machine’s closed ML system works from zero to train the particular job it will do. By taking a minimum spec chip that has already been created, the machine learns the process for creating that chip and continues to improve that process. These machines are not only able to detect problems but can figure out why these issues are occurring, decreasing production downtime.

Recently, ASML moved to Google Cloud. We talk about the products they use, why they chose Google Cloud, and their journey to the cloud.

Arnaud Hubaux

Arnaud Hubaux is development lead for Artificial Intelligence and Machine Learning products at ASML. He works with the world’s biggest chip manufacturers on AI-driven solutions to solve problems like optimizing production yield. As a trailblazer, Arnaud and his team implemented the first ASML AI application development pipeline on Google Cloud. With a PhD on applied constraint solving, Arnaud has deep experience as an architect of software platforms and product lifecycle management.

Cool things of the week

• Cloud migration: What you need to know (and where to find it) blog
• All together now: Fleet-wide monitoring for your Compute Engine VMs blog
• SRE Classroom: exercises for non-abstract large systems design blog
• gVisor: Protecting GKE and serverless users in the real world blog

Interview

• ASML site
• ASML in One Minute video
• Zoom in on the chip in your smartphone video
• KubeFlow site
• BigQuery site
• AI Platform Notebooks site
• Cloud Build site
• Introduction to Kubeflow video
• Intro to Kubeflow Pipelines video
• How to Build a Kubeflow Pipeline video

Tip of the week

We have a super cool tip on combining AI and WTH with meetings and webcams! Check out the Level Up - AI Director and Level Up - Real-Time Video Translation with AR Subtitles videos!

What’s something cool you’re working on?

Mark is working on some upcoming Kubernetes content.

Brian is working on an upcoming series of GCE videos with Carter Morgan.