• Bio
• Headshot
• Current talks
  • Machine Learning on Google Cloud
  • Using Kubeflow Pipelines for building machine learning pipelines
  • Visualizing Neural Networks with Activation Atlases
  • Getting started modeling language: a first steps in NLP with bag of words
• Consulting topics
  • Machine learning and data science infrastructure and architecture
  • Neural network architecture and design
  • Setting up your data pipeline for machine learning on GCP

Yufeng is a Developer Advocate at Google focusing on Cloud AI, where he is working to make machine learning more understandable and usable for all. He is the creator of the YouTube series AI Adventures, at yt.be/AIAdventures, exploring the art, science, and tools of machine learning. He enjoys hearing about new and interesting applications of machine learning, share your use case with him on Twitter @YufengG. He has previously spoken at many events around the globe, including Google I/O, Cloud Next, O'Reilly AI and Strata, PyCon, PyData, ML Prague, dotAI, and OSCON.

Twitter: @YufengG

LinkedIn: Yufeng Guo

Full-res headshot link in Drive

Here are some examples of talks I've given recently. New products, ideas, and use cases are always bubbling up to form new talks, so check with me to see if there's anything new and exciting coming up!

This interactive session will focus on the tools for doing machine learning on Google Cloud Platform (GCP). From exploration, to training, to model serving, we will talk about the available tools and how to piece them together depending on your needs. We will explore some common patterns, as well as leave some time to any address questions, so bring your use cases to discuss.

Kubeflow is an open-source project dedicated to making deployments of machine learning workflows on Kubernetes simple, portable and scalable. This session will focus on Kubeflow Pipelines, a platform to enable end-to-end orchestration of ML pipelines as well as easy experimentation and re-use. You'll learn how to build and manage machine learning workloads that can scale.

Kubeflow is a very exciting open-source project that bridges the gap between the DevOps world with the machine learning world. There are many concepts that can be highly valuable to cross-pollinate between these worlds, and Kubeflow helps codify that into best practices.

Learn more about Kubeflow Pipelines at https://www.kubeflow.org/docs/pipelines/pipelines-overview/

Neural networks work in "mysterious ways", but we can peer into them with more and more tools. This talk aims to show some interesting examples of different ways of visualizing neural networks with the goal of improving explainability. You will come away with a bit more understanding of how neural networks "learn", and what aspects they are "looking" at.

Freeform text contains a lot of rich data, but it can be hard to get at. Get a primer on natural language processing (NLP) and learn some of the tricks of trade. We'll talk about how to think about modeling language, and dive into the mechanics of how the bag-of-words model works.

These are open discussions and whiteboarding sessions where the customer comes away with a custom solution for their situation. I've presented a few examples below.

Having trouble deciding how to set up online and batch predictions without redundant work or bottlenecking your data? Trying to figure out how to set up your training infrastructure to give adequate flexibility without causing privacy issues? Something else keeping you up at night? Let's talk about your specific requirements, constraints, and wishes, to see what kind of cloud architecture would be most suitable.

Having trouble with getting those pesky models to converge? Can't quite get the kind of accuracy you were expecting? Let's chat about the intricacies of your data, model, and domain area to come up with a solution.

In an ever expanding world of data, there's more and more ways to mess up the data environment. From data leaking across orgs to missing or bottlenecked datasets, I'll help you work through the right customized solution for your setup.

The table of contents was generated using ekalinin's github-markdown-toc.