AWS Innovate Online Conference – AI and Machine Learning Edition


Hear the latest in machine learning from industry-leading scientists, AWS customers, and experts.

Thank you for attending the AWS Machine Learning Summit.
All content is now available on-demand.

Agenda at a Glance

60 mins Opening Keynote
45 mins Fireside chat
   Science of Machine Learning Impact of Machine Learning How Machine Learning is Done Machine Learning, No Expertise Required
30 mins Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100)
30 mins ipsum lorem (level 200) ipsum lorem (level 200) ipsum lorem (level 200) ipsum lorem (level 200)
30 mins Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100)
15 mins  Break
30 mins Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100)
30 mins ipsum lorem (level 200) ipsum lorem (level 200) ipsum lorem (level 200) ipsum lorem (level 200)
30 mins Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100) Lorem ipsum (level 100)

Select a track:

  • Science of Machine Learning
  • The ethical algorithm

    10:00 - 10:30 a.m.

    Recent mainstream media articles and books have raised alarms over antisocial algorithmic behavior, especially regarding machine learning (ML) and artificial intelligence. Concerns include leaks of sensitive personal data by predictive models, algorithmic discrimination as a side effect of ML, and inscrutable decisions made by complex models. However, an explosion of recent research in areas such as differential privacy, algorithmic fairness, and algorithmic game theory is forging a new science of socially aware algorithm design. This session surveys these developments and attempts to place them in a broader societal context. This session is largely based on the book “The Ethical Algorithm,” co-authored with Aaron Roth.

    Speaker(s): Michael Kearns, Amazon Scholar, AWS

    Towards controllable language generation

    10:45 - 11:15 a.m.

    Research in language generation has traditionally focused on representing constraints on choice and enabling content and word selection that conveys the desired speaker intent. The move to deep learning has brought dramatic improvements in the fluency and scalability of language generation systems but has sacrificed the ability to control choice. This session presents methods for faithful generation from data, faithful summarization of input text, and methods for controlling ordering of content. It also touches on a new summarization method under development, using an intermediate semantic representation, that has the potential to enable other types of fine-grained control of content and wording.

    Speaker(s): Kathleen McKeown, Amazon Scholar, AWS

    Analyzing social media for suicide risk using natural language processing

    11:30 - 12:00 p.m.

    The toll that mental illness takes worldwide is enormous. Since 2016, suicide has been the second leading cause of death in the US among ages 10–34 and the fourth among ages 35–54. COVID-19 has compounded this problem, as people have struggled with isolation, stress, and sustained disruptions to their lives. This session discusses opportunities and challenges in applying natural language processing and machine learning approaches to assess suicide risk and in computational research on mental health more generally. It includes a discussion of secure data enclaves as a practical way to make possible technological collaborations with sensitive data.

    Speaker(s): Philip Resnik, MLRA recipient, Professor of Computational Linguistics, University of Maryland

    Causality, robustness, and natural language understanding in ML

    12:15 - 12:45 p.m.

    AI is reaching many parts of our lives, from smart home devices using advanced natural language processing to computer vision. It also helps us understand our environments better and take better actions through causality and the robustness of systems equipped with it. In this Fireside Chat, join Alexa AI Sr. Principal Scientist Dilek Hakkani-Tur and Amazon Distinguished Scientists Alex Smola and Bernhard Schöelkopf for a discussion of current trends and what’s coming next.

    Speaker(s): Bernhard Schölkopf,  Amazon Distinguished Scientists, AWS
    Alex Smola Dilek, Amazon Distinguished Scientists, AWS
    Hakkani-Tur, Alexa AI Senior Principal Scientist, AWS

    Building high-quality computer vision models using only a few examples

    1:15 - 1:45 p.m.

    Business and other non-machine-learning-savvy users are increasingly needing to scale and automate tasks previously performed via visual inspection. Their problems are often specific and can’t be solved with general-purpose tools; their “labeled” data, from which a computer vision system can learn, is often scarce or expensive to gather. This requires learning from analogous experience, not just from data directly related to the problem. This session provides an overview of how the AWS Computer Vision group supports both users and the research community by offering suitable customer solutions—Amazon Rekognition Custom Labels and Amazon Lookout for Vision—and by publishing research.

    Speaker(s): Marzia Polito, Senior Manager, Applied Science. AWS

    Deep Graph Library: Deep Graph learning at scale

    2:00 - 2:30 p.m.

    Learning from graph and relational data plays a major role in many applications. In the last few years, graph neural networks (GNNs) have emerged as a promising supervised learning framework capable of bringing the power of deep representation learning to graph and relational data. This research has shown that GNNs achieve state-of-the-art performance for problems such as link prediction, fraud detection, knowledge-graph completion, and more. This session provides an overview of AWS AI/ML research in this area, which includes developing the Deep Graph Library (DGL)—an open-source framework for writing and training GNN-based models, scaling GNN model training to extremely large graphs, and developing GNN pretraining strategies.

    Speaker(s): George Karypis, Senior Principal Scientist, AWS

    COVIDcast: An ecosystem for COVID-19 tracking and forecasting

    2:45 - 3:15 p.m.

    In March 2020, the Delphi group at Carnegie Mellon University launched COVIDcast, which tracks and forecasts the spread of COVID-19. This project has many parts: (1) unique relationships with technology/healthcare partners granting access to data on pandemic activity; (2) infrastructure to build real-time, geographically detailed COVID-19 indicators from this data; (3) a historical database of all indicators, including revision tracking; (4) a public API serving new indicators daily (with R and Python client support); (5) interactive graphics to display indicators; and (6) forecasting and modeling work building on the indicators. This session provides an overview of COVIDcast, with demonstrations on how to access data and tools.

    Speaker(s): Ryan Tibshirani, Amazon Scholar, AWS
  • Impact of Machine Learning
  • Cultivating a company-wide machine learning culture at 3M

    10:00 - 10:30 a.m.

    3M leverages AWS technologies and services to create a new innovation space for scientists and engineers throughout the company. With innovation through new machine learning (ML) model development and advanced material and process technologies, developers create new value for customers. From reimagining research to streamlining manufacturing processes, 3M is building a new culture. In this session, learn how 3M is using their annual hackathon to grow their ML model development from one project to many in a short time.

    Speaker(s): David Frazee, Technical Director, 3M

    Transforming pharmaceutical R&D through industrialized machine learning

    10:45 - 11:15 a.m.

    Data and AI are central to AstraZeneca’s “Growth Through Innovation” strategy. The company has built an industrialized machine learning (ML) platform on AWS to streamline pharmaceutical drug discovery, clinical trials, and patient safety for hundreds of scientists. In this session, hear about AstraZeneca’s journey to build a flexible, compliant platform, how it helped to accelerate their drug research process, and how they collaborated with the Amazon Machine Learning Solutions Lab to make it happen.

    Speaker(s):  Dr. Ian Dix, Senior Director AI/ML and Analytics R&D,  AstraZeneca
    Rui Wang, Head of Compute and Core Engineering, R&D IT Data & Analytics, AstraZeneca

    ADP is designing ML solutions to deliver human capital insights

    11:30 - 12:00 p.m.

    As a leading global technology company supporting tens of millions of people with human capital management solutions, ADP manages an expansive collection of workforce data. The company wanted to leverage artificial intelligence and machine learning (ML) to help executives make real-time, data-informed decisions to better manage their businesses, while also adhering to the company’s commitment to privacy, model explainability, data governance, and ethical and practical data use. To better deliver these insights, ADP uses Amazon SageMaker to reduce their ML model deployment time from two weeks to one day. Learn how ADP uses data insights to deliver new value to customers and speed up innovation with Amazon SageMaker.

    Speaker(s): Jack Berkowitz, SVP of Product Development, ADP

    Bundesliga is delighting fans with game-changing insights

    12:15 - 12:45 p.m.

    Bundesliga, Germany’s premier national football league, uses machine learning (ML) and analytics to deliver real-time statistics to enrich the fan experience and recommend personalized match footage across mobile, online, streaming, and television broadcasts. In this session, learn how Bundesliga is changing the way fans watch professional soccer and get an inside look at how they created Bundesliga Match Facts like xGoals—the probability of a player scoring a goal when shooting from any position on the field.

    Speaker(s): Andreas Heyden, EVP of Digital Innovation, Bundesliga

    Vanguard is modernizing financial services through automation

    1:00 - 1:30 p.m.

    With over 30 million investors, over $6 trillion in assets, and more than 17 thousand employees, the Vanguard Group oversees a massive amount of financial information. The company set out to establish a secure and well-governed data science environment that provides transparency around data and model lineage, the model development process, model interpretability, and the performance of production models. In this session, learn how the company created a new, centralized machine learning (ML) platform to empower the company’s data science teams to innovate on behalf of customers more efficiently.

    Speaker(s): Ritesh Shah, Head of AI/ML Platforms w/in Chief Technology Office, Vanguard

    Fighting climate change at the edge with Carbon Lighthouse

    1:45 - 2:15 p.m.

    Commercial buildings are responsible for 40 percent of emissions in the US. Carbon Lighthouse is on a mission to stop climate change by making it easy and profitable for building owners to cut carbon emissions caused by wasted energy. In this session, learn how the company uses machine learning to develop insights that deliver energy savings and decrease CO2 emissions in commercial real estate, reducing more than 260,000 metric tons of emissions to date.

    Speaker(s): Brenden Millstein, President and Head of Product, Carbon Lighthouse
  • How Machine Learning is Done
  • Spin up Jupyter notebooks at scale and enhance your productivity

    09:45 - 10:15 a.m.

    Jupyter notebooks allow data scientists to create, share, and streamline work for data visualization from raw code to fully functional machine learning (ML) models. With Amazon SageMaker Studio, you can spin up Jupyter notebooks quickly without the need to manage the underlying compute resources. You can easily dial up or down the required compute scale, and the changes happen automatically in the background. The notebooks within SageMaker Studio are shareable, enabling collaboration with increased productivity. In this session, learn how you can use fully managed Jupyter notebooks and see an in-depth demonstration of building ML models at scale.

    Speaker(s): Brian Granger, Principal TPM, AI Platforms and Co-Founder, Jupyter

    Choose the right algorithm to help you build better ML models

    10:30 - 11:00 a.m.

    Amazon SageMaker offers a range of built-in machine learning (ML) algorithms for building and training ML models. With SageMaker, you can choose an ML algorithm specific to your data science requirements for both supervised and unsupervised learning. This session discusses the most popular algorithms used by data scientists using SageMaker. Dive deep into classification and regression with a k-nearest neighbors (k-NN) example as well as a demonstration of one of SageMaker’s many built-in computer vision (CV) algorithms. See these algorithms within the SageMaker Studio visual interface in an interactive demonstration.

    Speaker(s):  Denis Batalov , Tech Leader, AI & ML AWS

    Prepare data for machine learning with ease, speed, and accuracy

    11:15 - 11:45 a.m.

    Preparing data to build and train machine learning (ML) models is tedious. Data often resides in disparate sources and is not ready to use right away. Amazon SageMaker offers a range of solutions to better prepare your data for ML models. This session walks through a data preparation workflow in Amazon SageMaker. See how to connect multiple data sources and use prebuilt data visualization templates and data transforms to clean, verify, and explore data without writing code using SageMaker Data Wrangler. Learn how to use SageMaker Feature Store to create a repository to store, retrieve, and share features to enhance the efficiency of future model development.

    Speaker(s): Mark Roy, Principal ML Solutions Architect, AWS

    Tune your ML models to the highest accuracy with automatic model tuning

    12:00 - 12:30 p.m.

    Accuracy is critical in machine learning (ML), as we rely on correct predictions from ML models. To achieve accurate ML models, you need to tune arbitrary variables like learning rate and regularization to control the underlying algorithm, which requires a lot of trial-and-error experimentation. With automatic model tuning, Amazon SageMaker automatically tunes the relevant parameters and makes it easy to get the best possible outcomes for your ML models. Using Bayesian optimization techniques, SageMaker optimizes the values of the hyperparameters of your algorithm to achieve highly accurate ML models. This session demonstrates how you can achieve highly accurate models using automatic model tuning with SageMaker.

    Speaker(s): Emily Webber, Specialist, ML Solutions Architect, AWS

    Gain better insights while training ML models with SageMaker Debugger

    12:45 - 1:15 p.m.

    As state-of-the-art machine learning (ML) models grow in size and complexity, debugging becomes increasingly difficult. Just as unit tests boost traditional software development, an automated ML debugging library can save time and money. Amazon SageMaker Debugger is an ML feature that automatically identifies and stops underperforming training jobs. It automatically captures relevant data during training, such as model tensors, system resource utilization, and metrics about training operations. SageMaker Debugger provides rules that automatically identify issues related to model quality and computational performance. In this session, learn how you can profile and debug you training runs and how to set up real-time alerts.

    Speaker(s): Nathalie Rauschmayr, Applied Scientist, AWS Deep Engineering

    Accelerate NLP training with Amazon SageMaker

    1:30 - 2:00 p.m.

    The field of natural language processing (NLP) is developing rapidly, and NLP models are growing increasingly large and complex. Strong collaboration with organizations like Hugging Face and advanced distributed training capabilities make it easy to quickly train NLP models with Amazon SageMaker. In this session, learn how to quickly train an NLP model from the Hugging Face transformers library with just a few lines of code using PyTorch or TensorFlow and SageMaker’s distributed training libraries.

    Speaker(s): Aditya Bindal, Senior Product Manager, AWS Machine Learning

    Understand your ML models better with greater visibility and transparency

    2:15 - 2:45 p.m.

    Explaining machine learning (ML) models and understanding the reasoning behind the predictions they make is often difficult. Trained models may weigh some features more heavily than others when generating predictions. With Amazon SageMaker Clarify, you can develop a more robust understanding of your models by checking for potential bias in your training data or trained model, or by requesting explanations for model behavior. In this session, dive into the various bias metrics and importance graphs available in SageMaker Clarify across the ML workflow and discover how you can use them to improve your model.

    Speaker(s): Pinar Yilmaz, Senior Software Engineer, AWS Machine Learning

    Accelerate deep learning in the cloud with custom ML environments

    3:00 - 3:30 p.m.

    Deep learning (DL) projects often require integrating custom or domain-specific libraries with popular open-source frameworks such as TensorFlow and PyTorch. Setting up, managing, and scaling custom machine learning (ML) environments can be time-consuming and cumbersome, even for experienced ML practitioners. With AWS Deep Learning AMIs (DLAMI) and AWS Deep Learning Containers, you get access to prepackaged and optimized DL frameworks that make it easy for you to customize, extend, and scale your environments. In this session, learn how to launch and use DLAMI; how to pull, customize, and extend Deep Learning Containers; and how to run large-scale training experiments with Amazon EKS and Amazon SageMaker.

    Speaker(s): Shashank Prasanna, Senior Developer Advocate, AI & ML, AWS
  • Machine Learning, No Expertise Required
  • Automate your document pipeline with intelligent document processing

    9:00 - 9:30 a.m.

    You have documents—tons of them. You have customers who expect quick responses and outcomes based on the information in those documents. Using AWS machine learning, you can implement a document processing solution, and this session shows you how. Using Amazon Textract to extract data, Amazon Comprehend to pull out custom entities, and Amazon Augmented AI (Amazon A2I) to review the results, you can process financial documents faster than ever and at scale. Join us as we walk you through how to set up a document processing pipeline that saves time, reduces error, and enables better employee productivity.

    Speaker(s): Prem Ranga, Senior Solutions Architecture, AWS

    Give users relevant answers quickly with intelligent search

    9:45 - 10:15 a.m.

    Search relevance isn't simply about the retrieval of content. It’s about better understanding user intent so accurate information is prioritized, and that’s how you empower employees and delight customers. In this session, see how ML-powered intelligent search can be used to provide the right information to the right person at the right time, whether that’s role-relevant material within the workplace or quality curated content for end consumers.

    Speaker(s):  Sara van de Moosdijk, Senior AI/ML Partner SA, AWS

    Transforming DevOps at Fidelity Investments with intelligent insights

    10:30 - 11:00 a.m.

    While DevOps-style tool chains and mechanisms have helped shorten response time, IT teams are still overwhelmed with the volume of data and activities required. There is a growing need for a next-generation model that integrates development and operations cycles with automated practices to operate efficiently. Join this session to learn how Fidelity Investments leverages Amazon DevOps Guru and its ML-powered insights to improve developer operations and the availability of their applications.

    Speaker(s): Jacob Sullivan, Senior Manager, Product Management, AWS
    Keith Blizard, SVP SRE & Observability, Fidelity

    Transform online shopping experiences with personalized recommendations

    11:15 - 11:45 a.m.

    Customers today expect a highly personalized journey when shopping, and by using machine learning, brands have the power to turn every shopping interaction into a meaningful and rewarding experience. In this session, learn how you can help customers quickly find the products they’re searching for using personalized product ranking and recommendations powered by Amazon Personalize.

    Speaker(s): Sarah Strobhar, Senior AI/ML GTM Specialist, AWS

    Increase productivity and satisfaction with an intelligent contact center

    12:00 - 12:30 p.m.

    Adding intelligence to your call center can help you better understand your customer and improve your overall service ratings. AWS and AWS Partners provide a broad set of integrated solutions powered by ML so you can build seamless customer experiences that reduce operational costs, increase agent productivity, and improve customer satisfaction. In this session, learn how Amazon Connect and AWS Contact Center Intelligence (CCI) solutions can modernize your contact center and drive operational advances on the contact center system of your choice.

    Speaker(s):  Shanthan Kesharaju Global Solutions Lead
    Dave Rennyson CEO, SuccessKPI

    Get started with machine learning with AWS DeepRacer

    12:45 - 1:15 p.m.

    Developers, start your engines! This session provides developers of all skill levels an opportunity to get hands-on experience with AWS DeepRacer and hear about exciting announcements and enhancements coming to the League in 2021. Learn about the basics of machine learning (ML) and reinforcement learning—an ML technique ideal for autonomous driving. In this session, you can build a reinforcement learning model and submit it to the AWS DeepRacer League for a chance to win prizes and glory.

    Speaker(s): Naomi Teng, Innovation Program Manager, AWS

    Transform your business with data-driven insights

    1:30 - 2:00 p.m.

    Business metric data is a treasure trove of insights. Getting to those insights requires either a lot of time or the right tools. The predictive nature of machine learning (ML) can leverage your data to provide proactive, actionable insights that enable you to act on new opportunities, catch issues before they occur, and better anticipate your business needs. In this session, learn how to apply ML using Amazon Forecast, Amazon Fraud Detector, and Amazon Lookout for Metrics to help you accurately predict supply/demand forecasting, find online fraud before it happens, detect and diagnose anomalies in business and operational data, and more—all with no ML expertise required.

    Speaker(s): Karl Albertsen , Senior Product Manager, AWS

    Bring AutoML to all users across your business

    2:15 - 2:45 p.m.

    Join this session to learn how to get started quickly with AutoML from AWS. We offer over 70 solutions and services with AutoML built in, and our unique approach makes AutoML available to everyone across your business—from data scientists who want to explore model creation to app developers looking to apply AI—with no machine learning expertise required. This session explains how AWS built AutoML capabilities into our services, including Amazon SageMaker Autopilot, Amazon Rekognition, and Amazon Comprehend. If you are curious about AutoML, this session is for you!

    Speaker(s): Kate Zimmerman, Principal Deep Learning Architect, AWS

    Maximize lifetime value of media content with AWS Media Intelligence

    3:00 - 3:30 p.m.

    Media content is being created faster than ever before, but the process to produce high-quality, localized content is often complex and expensive. Finding ways to efficiently moderate and monetize content at scale is also difficult. Join this session to learn how AWS Media Intelligence solutions powered by AI services help with content search and discovery, captioning and localization, compliance, and content monetization while lowering operational costs and increasing content value.

    Speaker(s):  Esther Lee, Senior Product Manager, AWS
    James Jameson, Head of WW M&E data science, Caption Hub

Level 100

Sessions focus on providing an overview of AWS services and features, with the assumption that attendees are new to the topic.

Level 200

Sessions focus on providing best practices, details of service features, and demos with the assumption that attendees have introductory knowledge of the topics.

Level 300

Sessions dive deeper into the selected topic. Presenters assume that the audience has some familiarity with the topic but may or may not have direct experience implementing a similar solution.