Agenda

Welcoming Remarks &
Overiew of Google Developers Program in North America

About Kyle

Kyle works for Google in Mountain View, California. His job on the Developer Relations team is to support awesome developer communities, like the Google Developer Experts and Developer Student Clubs. In his spare time he enjoys building and hacking on the web, playing with his two cats, and photographing the outdoors. Before Google, Kyle was a startup founder, organizer for GDG Kansas City, and a Google Developer Expert (GDE).

Time-Series Forecasting using the Google Cloud AI Platform

Predicting the future has always been a fascinating topic. Now we have AI tools and techniques that can help us do it better than ever before. In this session, we'll cover the fundamentals of solving time-series problems with AI, and show how it can be done with popular data science tools such as Pandas, TensorFlow, and the Google Cloud AI Platform. We'll start with how to visualize, transform, and split time-series data for use in an ML model. We'll also discuss both statistical and machine learning techniques for predictive analytics. Finally, we'll show how to train a demand forecasting model in the cloud and make predictions with it. Attendees can access Jupyter notebooks after the session to review the material in more detail.

About Karl

Karl Weinmeister is a Cloud AI Advocacy Manager at Google, where he leads a team of data science experts who develop content and engage with communities worldwide. Karl has worked extensively in machine learning and cloud technologies. He was a contributor to one of the first AI-based crossword puzzle solvers that is still referenced today.

Scaling your TFX Training Pipeline using TPU Pods

TFX is a Google-production-scale machine learning (ML) platform based on TensorFlow. Scaling your TFX Training Pipeline is a must for certain use-cases and industry scenarios. Traditionally, ML Engineers on GCP can use TPUs on Cloud TPU VM instance cluster of workers. In this talk, I will show how to scale your TFX Training pipeline using TPU Pods v2 and v3 via AI Platform Jobs. AI Platform Jobs promotes an on-demand serverless approach for managing training tasks, costs and resources consumption.

About Carlos

Carlos is a Machine Learning Engineer at Pythian. He attains Cloud Architect, Data Engineer Google Cloud Professional certifications. Also, he is AI/ML Consultant to SMB in the Americas. Nowadays, he builds and deploys ML system in production using Google Cloud Platform. As GDE in ML and GDG Cloud organizer, he trains PhD students in Brazil and Portugal on Data Science.

Intermission / Break

Driving better user experience via the reward function of RL-based recommenders

Can we design recommenders that encourage user trajectories aligned with the true underlying user utilities? Besides engagement, user satisfaction and responsibility arise as important pillars of the recommendation problem. Motivated by this, we will discuss various efforts utilizing the reward function as an important lever of Reinforcement Learning (RL)-based recommenders, so to guide the model learning that for certain states (i.e., latent user representation at a certain point of the trajectory) certain actions (i.e., items to recommend) will bring higher user utility than others. We will also outline current and future directions on overcoming challenges of signals’ sparsity and interplay among various reward signals.

About Konstantina

I am a Research Engineer at Google, leading several efforts on recommender systems and reinforcement learning in Google Brain. I am leading research and product engagements applying Reinforcement Learning to increase user satisfaction with Google products as well as make the systems socially responsible.

ML Engineering for Production ML Deployments

Delivering the results of advanced Machine Learning technology to customers requires a rigorous approach and production-ready systems. This is especially true for maintaining and improving model performance over the lifetime of a production application. Unfortunately, the issues involved and approaches available are often poorly understood. An ML application in production must address all of the issues of modern software development methodology, as well as issues unique to ML and data science. Often ML applications are developed using tools and systems which suffer from inherent limitations in testability, scalability across clusters, training/serving skew, and the modularity and reusability of components. In addition, ML application measurement often emphasizes top level metrics, leading to issues in model fairness as well as predictive performance across user segments. We discuss the use of ML pipeline architectures for implementing production ML applications, and in particular we review Google’s experience with TensorFlow Extended (TFX), as well as the advantages of containerizing pipeline architectures using platforms such as Kubeflow. Google uses TFX for large scale ML applications, and offers an open-source version to the community. TFX scales to very large training sets and very high request volumes, and enables strong software methodology including testability, hot versioning, and deep performance analysis.

About Robert

A data scientist and TensorFlow addict, Robert has a passion for helping developers quickly learn what they need to be productive. He's used TensorFlow since the very early days and is excited about how it's evolving quickly to become even better than it already is. Before moving to data science Robert led software engineering teams for both large and small companies, always focusing on clean, elegant solutions to well-defined needs. You can find him on Twitter at @robert_crowe.

Machine Learning: Image recognition on the Raspberry Pi

Are you a hobbyist that wants to add some machine learning to your projects? Are you a machine learning engineer and want to add some hardware to your project? Are you just interested in either? This talk is for you. I will walk through the process of setting up a Raspberry Pi, connecting a camera, adding the proper machine learning libraries (TensorFlow and Keras), and then implementing a pre-trained image recognition network. When it is all done, you will be able to use your Pi to detect what object it is seeing.

About Evan

Evan Hennis is a Google Developer Expert in Machine Learning and an international speaker. He has a Master's degree in Computer Science with a specialization in machine learning from Georgia Tech. He can be reached on Twitter at @TheNurl or followed on his blog at http://blog.eckronsoftware.com/.

Introduction to Swift for Tensorflow

We will use Swift for TensorFlow to build a simple neural network and use it to categorize MNIST digits, then look at how we can extend our approach to run on different hardware using Google's cloud. Along the way, we will look at how Swift works with the LLVM compiler and automatic differentiation to make it easier to reason about our code.

About Brett

Brett Koonce is the CTO/co-founder of Quarkworks, a mobile consulting agency. He has worked on dozens of apps and contributed code to many different open source projects. He enjoys building and scaling teams to solve interesting problems. His upcoming book "Convolutional Neural Networks with Swift for TensorFlow" is available for preorder from Apress.

Closing Remarks

Welcoming Remarks &
Overiew of Google Developers Program in North America

About Kyle

Kyle works for Google in Mountain View, California. His job on the Developer Relations team is to support awesome developer communities, like the Google Developer Experts and Developer Student Clubs. In his spare time he enjoys building and hacking on the web, playing with his two cats, and photographing the outdoors. Before Google, Kyle was a startup founder, organizer for GDG Kansas City, and a Google Developer Expert (GDE).

Federated Learning with TensorFlow Federated

Meet federated learning: a privacy-preserving machine learning technology for data that is distributed across a fleet of compute nodes, such as cell phones, smart cars, or data driven hospitals. With federated learning, clients collaboratively train a model under the direction of a server, while keeping the training data decentralized. Learn how you can use TensorFlow Federated to explore federated learning.

About Emily

Emily is a software engineer on Google's federated learning team. She studied Electrical Engineering at the University of Iowa prior coming to Google.

About Chunxiang (Jake)

Chunxiang graduated from University of Washington with a Ph.D in Chemistry, currently a software engineer at Google AI.

MLOps and Metadata - Why your ML projects needs both

Current machine learning projects involve a variety of manual steps performed by data scientists, e.g. data or model validation. Those tasks prevent the data scientists from developing new models and innovating solutions to business problems. MLOps is solving this challenge by providing ML pipeline concepts and tools to continuously update and deploy machine learning models. During the pipeline process, metadata is produced by each pipeline steps which can further aid the data scientists, ,e.g. by tracking the changes of models. In this talk, Hannes is introducing the concept of machine learning pipelines and how the machine learning metadata can be used in data science projects. Both concepts will reduce the burden of data scientists and free their valuable time to focus on new data science problems.

About Hannes

Hannes Hapke is a senior machine learning engineer for Concur Labs at SAP Concur, where he explores innovative ways to use machine learning to improve the experience of a business traveler. Prior to joining SAP Concur, Hannes solved machine learning problems in various industries including healthcare, retail, recruiting, and renewable energies. He has co-authored machine learning publications including O’Reilly’s presentation on "Building Machine Learning Pipeline" and Manning’s publication “Natural Language Processing in Action”. At SAP, Hannes is focusing on Machine Learning Engineering and NLP problems.

Intermission / Break

What Does a Model Learn?

How can we explain how deep neural networks arrive at decisions? Feature representation is complex and to the human eye opaque; instead a set of interpretability tools intuit what the model has learned by looking at what inputs it pays attention to. This talk will introduce some of the challenges associated with interpretability for deep neural networks and discuss desirable properties methods should fulfill in order to build trust between humans and algorithms.

About Sara

Sara Hooker is a researcher at Google Brain doing deep learning research on reliable explanations of model predictions for black-box models. Her main research interests gravitate towards interpretability, model compression and security. In 2014, she founded Delta Analytics, a non-profit dedicated to bringing technical capacity to help non-profits across the world use machine learning for good. She grew up in Africa, in Mozambique, Lesotho, Swaziland, South Africa, and Kenya.

Using TensorFlow on Google Cloud Platform

Whether you've built machine learning models with TensorFlow, or if you are just getting started, this session will show you how to effectively use TensorFlow in the Cloud. We'll provide hands-on tips and code samples for each step of the process. We will show how to use notebooks, pre-configured VM images, serverless training and serving tools, explanations, TPUs, and pipelines.

About Karl

Karl Weinmeister is a Cloud AI Advocacy Manager at Google, where he leads a team of data science experts who develop content and engage with communities worldwide. Karl has worked extensively in machine learning and cloud technologies. He was a contributor to one of the first AI-based crossword puzzle solvers that is still referenced today.

Productionizing ML with ML Ops and Cloud AI

The hardest part of ML adoption in enterprises is Productinization. As we see in recent discussions around ML Ops, there is a big gap between Data Scientists' PoC code and production ML development and operation with the Ops team. Such as, preparing a manageable ML dev environment, building a scalable ML serving infrastructure, setting up a ML pipeline for continuous training, and automated validation of data and model. In this session, we will learn how to leverage various Google's ML/AI offerings such as TensorFlow Extension (TFX), TensorFlow Enterprise, Cloud AI Platform Notebooks, Training, Prediction, and Pipelines for productionizing your ML service with the ML Ops best practices.

About Kaz

Kaz Sato is Staff Developer Advocate at Google Cloud for machine learning and AI products, such as TensorFlow, Cloud AI and BigQuery. Kaz has been invited as a speaker at major events including Google Cloud Next, Google I/O, NVIDIA GTC and etc. Also, authoring many GCP blog posts, supporting developer communities for Google Cloud for over 9 years. He is also interested in hardwares and IoT, and has been hosting FPGA meetups since 2013.

Closing Remarks

Welcoming Remarks &
Overiew of Google Developers Program in North America

About Kyle

Kyle works for Google in Mountain View, California. His job on the Developer Relations team is to support awesome developer communities, like the Google Developer Experts and Developer Student Clubs. In his spare time he enjoys building and hacking on the web, playing with his two cats, and photographing the outdoors. Before Google, Kyle was a startup founder, organizer for GDG Kansas City, and a Google Developer Expert (GDE).

Full Cycle Data Science w/ GCP

Can a single data scientist truly make an impact on their product? Can a single data scientist own the full product cycle from research of a vague idea into owning a production-grade ML model inference service? Can that data scientist do all of that with purely ML & simple Python expertise?! YES! Not only its possible, I will claim that this is THE ONLY WAY an organization can scale its data science based products and become purely intelligent. This dreamy reality is only possible with a 2020+ like ML infrastructure. Luckily GCP offers such an infrastructure I will describe the full cycle data science philosophy and architecture on GCP, along with our journey (which is never ending) and success stories from doing that at Waze.

About Daniel

Tech leading data science & data architecture (office of the CTO) @Google, Waze where in the last 6 years I’ve had a bunch of different titles including data scientist, data engineer, data architect, etc. During that time I got to lead cool projects like Carpool matching (ML model), ETA & routing for motorcycles (ML model), funnel / behavioural self service analytics platform (not an ML model but still fun).

Humans vs. AutoML: How to improve your ML models with BigQuery ML

How do human data scientists compare to AutoML Tables? I'll demonstrate several approaches to how humans can improve their machine learning model in BigQuery ML, but stay tuned to find out whether human data scientists or AutoML win the race towards the highest AUC score in a Kaggle Competition!

About Polong

Polong helps data scientists make the most of Google Cloud. He has taught data science and machine learning to hundreds of thousands of learners online.

Intermission / Break

Using Google’s Cloud AI APIs To Watch, Visualize And Forecast The World In Realtime

The GDELT Project is one of the largest open datasets for understanding human society, totaling more than 3.2 trillion datapoints spanning 200 years in 152 languages. From mapping global conflict and modeling global narratives to providing the data behind one of the earliest alerts of the COVID-19 pandemic, GDELT explores how we can use data to let us see the world through the eyes of others and even forecast the future, capturing the realtime heartbeat of the planet we call home. What does it look like to analyze, visualize and even forecast the world in realtime through the eyes of GCP’s vast array of AI and analytic offerings, from sampling billions of news articles through the Natural Language API to cataloging half a billion images through Cloud Vision to watching a decade of television news with Cloud Video? From visual search to misinformation research to planetary-scale semantic and visual analysis, what does it look like to analyze the global news landscape through the eyes of today’s cloud AI and how does one transform the resulting vast archives of JSON annotations into actionable insights using tools like BigQuery and Inference API?

About Kalev

Dr. Kalev Hannes Leetaru - One of Foreign Policy Magazine's Top 100 Global Thinkers of 2013, Kalev founded the open data GDELT Project. From 2013-2014 he was the Yahoo! Fellow in Residence of International Values, Communications Technology & the Global Internet at Georgetown University's Edmund A. Walsh School of Foreign Service, where he was also an Adjunct Assistant Professor, as well as a Council Member of the World Economic Forum's Global Agenda Council on the Future of Government. His work has been profiled in the presses of more than 100 nations and in 2011 The Economist selected his Culturomics 2.0 study as one of just five science discoveries deemed the most significant developments of 2011. Kalev’s work focuses on how innovative applications of the world's largest datasets, computing platforms, algorithms and mind-sets can reimagine the way we understand and interact with our global world. More on his latest projects can be found on his website at https://www.kalevleetaru.com/ or https://blog.gdeltproject.org.

Data Science at Scale Using R on Google Cloud

R is a very popular programming language in the data science world, with a vibrant community of developers continuing to increase its power and versatility. Come learn how Google Cloud can help data scientists get even more out of R in their day-to-day work, using AI Platform Notebooks. We’ll demonstrate how these R notebooks require minimal configuration and can be set up with high memory, allowing analysis of much bigger datasets and on a larger scale than is typically done on a single computer instance of R. Other features of AI Platform Notebooks that will be highlighted include integration with other Google Cloud tools, interactivity and reproducibility - each key in its own way to the value of many R-based analyses.

About Alok

Alok is a Data Science Advocate at Google, where he shows how to use Google Cloud tools for data science, in sports and otherwise. He is an experienced and versatile data scientist with strong work ethic, collaborative mindset, and particular expertise in sports analytics.

Building Scalable Machine Learning on Medical Records

In this talk, we provide a practical tour of major methodological approaches in building analytics and machine learning systems in the healthcare domain . Utilizing Jupyter notebook and synthetic FHIR data, we will demo nuances of working with healthcare data and methods to convert leagcy data representations machine learning ready dataset. Lastly, demonstrate a reproducible ML pipeline- training/testing/validation data creation, feature engineering, evaluation metrics, and model deployments using Google Cloud AI Platform tools.

About Adrish

Adrish Sannyasi is a senior level Healthcare Solutions Manager and Customer Success leader with Google Cloud- with a background in software engineering and clinical informatics/data science; focuses on solving problems in healthcare delivery and helping provider and payor organizations on improving customer experiences through healthcare analytics and AI applications in secure cloud infrastructure and modern API(s). Adrish obtained a BS in electrical engineering, graduate certificates in Biomedical Informatics (OHSU), and Medical Data Modeling and Analysis (Stanford School of Medicine), and an MBA in Operations Management from the University of Maryland. Prior to joining Google in 2018, he worked at Machine Intelligence startup Ayasdi as a healthcare data scientist. Before Ayasdi, he held sales engineering/healthcare analytics specialist roles at Splunk and Oracle Corporation and worked as a healthcare IT technical consultant at Deloitte Consulting. 

Closing Remarks