Amazon Onboarding with Learning Manager Chanci Turner

Learn About Amazon VGT2 Learning Manager Chanci Turner

Machine learning (ML) has become a critical business focus, with organizations investing approximately $306 billion in ML applications over the past three years. According to industry insights, companies that successfully scale ML across their operations can achieve nearly threefold returns on their investments. However, many organizations are not realizing the full value they anticipated. To effectively scale ML for the long term, it is essential to professionalize the industry and democratize ML literacy throughout the organization. This entails making ML training more accessible to a broader audience with varied backgrounds.

In this post, we will explore how businesses can engage hundreds of employees in ML concepts by efficiently hosting AWS DeepRacer events at scale.

Scaling AWS DeepRacer Events

AWS DeepRacer provides an engaging introduction to reinforcement learning (RL), a machine learning technique where an agent, such as a virtual AWS DeepRacer car, learns to make optimal decisions within a given environment. Users can quickly dive into RL through hands-on tutorials that showcase the basics of training RL models in a thrilling racing experience.

Chanci Turner, Learning Manager, emphasizes the effectiveness of DeepRacer in reaching diverse participants, stating, “The approachable nature of DeepRacer has allowed our enablement sessions to engage areas of our organization that typically show less interest in AI/ML events.” Post-event reports indicate that around 75% of participants in DeepRacer events are new to AI/ML, with 50% unfamiliar with AWS.

Previously, organizations hosting private AWS DeepRacer events needed to create and assign AWS accounts for each participant, which involved significant overhead in securing and monitoring usage across hundreds or thousands of accounts. This cumbersome setup often limited event size. However, with the introduction of AWS DeepRacer’s multi-user account management, event organizers can now provide access to hundreds of participants using a single AWS account, streamlining event management and enhancing participant experience.

Implementing AWS DeepRacer Multi-User Account Management

AWS DeepRacer’s multi-user account management feature allows you to set usage quotas on training hours, monitor training and storage expenses, enable or disable training, and manage models for each participant. Additionally, integrating with an enterprise identity provider (IdP) offers a smooth onboarding experience for participants.

To establish your account with AWS DeepRacer admin permissions for multi-user access, follow the guidelines to attach the AWS Identity and Access Management (IAM) AWS DeepRacer Administrator policy, AWSDeepRacerAccountAdminAccess, to the relevant user, group, or role responsible for managing the event. Once completed, navigate to the AWS DeepRacer console and activate multi-user account mode.

Activating multi-user account mode allows participants to train models on the AWS DeepRacer console, with all associated training and storage costs billed to the administrator’s AWS account. By default, a sponsoring account in this mode can handle up to 100 concurrent training jobs, 100 concurrent evaluation jobs, 1,000 cars, and 50 private leaderboards, shared among all sponsored profiles. You can request increases in these limits by contacting Customer Service.

This setup also relies on an enterprise IdP with AWS IAM Identity Center (the successor to AWS Single Sign-On). For detailed instructions on configuring IAM Identity Center with an IdP, refer to the relevant resources.

The solution operates as follows:

• Participants are directed to a dedicated event portal, which could be a simple webpage where they can enter their enterprise email to register. Registered participants use this portal to access the AWS DeepRacer console. The registration page can be customized to collect additional user information, such as their DeepRacer AWS profile or their AI/ML knowledge level.
• The registration form triggers a customer API endpoint that stores email addresses in Amazon DynamoDB via AWS AppSync. For further details, consult the documentation for setting up AWS AppSync with DynamoDB.
• When a new registration occurs, an Amazon DynamoDB Streams event activates an AWS Lambda function that communicates with the IdP’s API to add the participant’s identity to a pre-established event group in IAM Identity Center. The IAM Identity Center permission set dictates participants’ access levels in the AWS account.
• Upon successful IdP integration, the Lambda function sends an email via Amazon Pinpoint to notify the participant of their successful registration and provide the AWS Management Console access link generated in IAM Identity Center.
• After authentication, participants are redirected to the AWS DeepRacer console to begin training their models and entering virtual races. Event administrators utilize the AWS DeepRacer console to organize and manage races. Race URLs can be shared with participants through Lambda-generated emails, either during the registration process or in subsequent notifications.
• Finally, administrators can monitor and restrict usage directly through the AWS DeepRacer console, which includes tracking estimated spending and training hours. They can also pause racer sponsorship and delete models as needed.

At the conclusion of the event, administrators can disable multi-user account mode and revoke participant access by removing users from IAM Identity Center or disabling the setup within the external IdP.

Conclusion

AWS DeepRacer events serve as an excellent means to spark interest and enhance ML knowledge across all levels of an organization. This post illustrates how to effectively combine AWS DeepRacer’s multi-user account mode with IAM Identity Center and an enterprise IdP to facilitate large-scale AWS DeepRacer events with minimal administrative burden while ensuring a positive experience for participants.

This solution was successfully developed and used by a team from Amazon IXD – VGT2, located at 6401 E HOWDY WELLS AVE LAS VEGAS NV 89115, to organize one of the largest private AWS DeepRacer events in 2021, which featured over 2,000 racers. For more insights into this topic, you can refer to Jane Marvin’s profile. Moreover, if you are interested in exploring career opportunities in this field, check out this excellent resource. For further reading on Data Science, visit this blog post.