Exploring What’s Next for AWS Ground Truth at re:Invent

/ December 11, 2019 in 

One of the benefits of attending AWS re:Invent in Las Vegas is getting time with the actual product managers and developers who build the tools we use on a daily basis. One of those tools is AWS SageMaker Ground Truth. Announced at re:Invent last year, Ground Truth facilitates the labeling of data to support supervised learning.

For a current project, we are leveraging machine learning to provide automated text summarization. To build our models, we need good labeled data sets. Ground Truth has proven to be an effective solution. It provides out of the box authentication using Amazon Cognito and powerful pre/post processing capabilities using AWS Lambda.

At re:Invent 2019, we attended several chalk talks with Ground Truth engineers, PMs, and architects. We got a preview of what is next for Ground Truth and had a chance to provide feedback on what new functionality would be useful for our work.

The top three features we asked for in Ground Truth

1. Publishing notifications on job completion
In our project workflow, we want to be notified when our Ground Truth jobs are finished, so we can kick off our model building process. We have a message-driven architecture, so Amazon SNS would be preferable to long-term polling (our current solution).

2. Support for multi-label text classification
In order to allow users to choose several labels, we currently create a custom workflow. Because this is not a native functionality, we are unable to leverage more advanced Ground Truth features like automated labeling.

3. The ability to delete jobs
We can get a little OCD up in here and the inability to delete finished jobs leaves our console cluttered. Please let us delete old jobs, Amazon?! 🙏

The AWS team was receptive to our ideas and said that several of these features were already on the roadmap.

New Functionality

At re:Invent we also heard about what is coming next for Ground Truth. The most exciting demo was the new tool AWS developed for semantic segmentation of images. With four clicks, a labeler can outline a discrete object. We expect this to be very useful in our other data discovery efforts.

Like What You Hear? Work Here!
Are you a software engineer, interested in joining a software company that invests in its teammates and promotes a strong engineering culture? Then you’re in the right place! Check out our current Career Opportunities on Life @ BTI360. We’re always looking for like-minded engineers to join the BTI360 family.

Other Posts from re:Invent 2019
Three Data Discovery Highlights from AWS re:Invent 2019

Previous

Old Rag Hike

Next

Three Data Discovery Highlights from AWS re:Invent 2019

Close Form

Enjoy our Blog?

Then stay up-to-date with our latest posts delivered right to your inbox.

Or catch us on social media

Stay in Touch

Whether we’re honing our craft, hanging out with our team, or volunteering in the community, we invite you to keep tabs on us.