#TFX supports Apache Airflow and Kubeflow orchestration, we designed it for production TensorFlow workloads tensorflow.org/tfx #TFDevSummit

A great way to get started with TensorFlow Extended (#TFX) is the tutorial just published by Robert Crowe :) tensorflow.org/tfx/tutorials/…

You can now create a complete TensorFlow Extended (TFX) pipeline using several ready-made components, configure them for many ML use cases, and execute with an orchestration system of your choice. Read Jarek Wilkiewicz’s blog post to learn more ↓ goo.gle/2QnQfvp

Honored to be an invited speaker at the #VLDB2019 panel (w/Alkis Polyzotis Theo Rekatsinas Ce Zhang & Sudeepa Roy) on DB + ML research! I will attempt to draw some obvious, some not obvious, and some controversial lessons from #ModelDB and Verta! Pre-panel questions/ideas welcome.

Great to have Alkis Polyzotis of Google AI visit and give a super interesting and eye-opening talk on data-centric issues in production ML platforms in my course on data+ML+systems for data science majors! cseweb.ucsd.edu/~arunkk/dsc102… @DataSciUCSD UCSD CSE UCSD Engineering

“The narrative has changed. ML projects at Google now focus in data validation *before* they lay down production pipelines”. Alkis Polyzotis of Google TFX team talking about data quality for ML at Berkeley.

"Just like you wouldn’t code without version control, you shouldn’t train models without lineage tracking." Couldn't agree more with this message! Great launch for Cloud AI Pipelines, building on our project on ML Metadata. cloud.google.com/blog/products/…

Super excited to see this toolkit released, along with its integration with MLMD and TFX/OSS. Yet another example of how data/model/metrics/${artifact} provenance becomes important in ML workflows.

Coping with Rejection in Academia: A Three-Pronged Approach. loom.com/share/89bfb106…

Framework coming out of my team at Google. Please take a look! Very proud of this first release, and many more improvements planned.

Together, we will build a world where the accomplishments of our daughters will be celebrated, rather than diminished.

📁 Track where your models came from with ML Metadata ML Metadata is a library for TensorFlow and other ML frameworks that lets you record and query the full lineage of your models. And it comes integrated within TensorFlow Extended (TFX). Learn more ↓ goo.gle/3i32r2A

Great to see the years of data exchange work culminating with the Church Award. Today, at CSL'21, with my colleagues Ron Fagin, Phokion Kolaitis, Renée Miller renee miller and Wang-Chiew Tan Wang-Chiew Tan Great talk by Phokion! youtube.com/watch?v=Qk0GdY…

Very excited about the launch of Databricks ML! Many companies offer ML platforms, but they’re always different from data platforms, which makes no sense because ML is all about data (ML = algorithms that extrapolate from data). databricks.com/blog/2021/05/2…

Lots of nice content at the NeurIPS #DataCentricAI workshop today. I had fun writing this paper with Alkis Polyzotis on what Data-Centric AI can learn from Data and ML engineering. There's a lot needed to make AI not just data-centric but production-ready. arxiv.org/pdf/2112.06439…

Is there less intrinsic severity of Omicron than Delta and prior variants? Update from today's UK report gov.uk/government/pub… Less frequent loss of smell/taste and in-lab lung cell infectivity (#6 below). (Still too early to know for sure)

Apache Spark has won the SIGMOD Systems Award, and Photon won the Best Industry Paper Award! This is a great testament to the adoption and impact of these two systems. If you are at SIGMOD this wk, please come by and say hi. databricks.com/blog/2022/06/1…

I had a lot of fun preparing for the Data Platform Summit keynotes tomorrow! Tune in at databricks.com/dataaisummit -- online access is free. We'll be presenting a ton of new efforts across Apache Spark, Delta Lake, MLflow, #DeltaSharing, and the Databricks platform.

Databricks Model Serving is now GA! Key benefits: - Simple: pick a model and get HA serving and fast autoscaling to minimize cost. - Integrated with your data platform: build features, serve them, monitor and debug quality all on the same platform. databricks.com/blog/2023/03/0…