Amazon Web Services’ (AWS) re:Invent this year is a little different. There’s no week-long conference in Las Vegas where attendees scurry between keynotes, leadership sessions, lightning talks, and Deep Racer. It is a completely virtual conference that will run for 3 weeks where attendees can tune in from the comfort of their couches. Even Deep Racer is being played online. With almost all sessions being recorded and re-broadcast there seems to be way too much for anyone to digest. For free!
The conference fever over at PREDICTif was amplified under a whole other context: re:Invent 2020 added a new keynote just for machine learning! Machine learning is beginning to hog the headlines and that could not sound any better to the PREDICTif geeks who play in this space for a living.
The AWS machine learning platform, Amazon SageMaker, saw a bunch of new features added – Pipelines, Edge Manager, Feature Store & Data Wrangler. Not only did we see the native platform grow, but also the Machine Learning concept being peppered into other key AWS services like Amazon RedShift and Amazon QuickSight.
a fine frenzy
The biggest frenzy thus far was caused by Amazon SageMaker Clarify–a gamechanging feature added to SageMaker. Clarify is an open-source library that helps detect bias in ML systems. Like a helper checker, Clarify attempts to de-risk the biases of Data Scientists who train and deploy these models. Using statistical analysis, it helps the modeler identify imbalances while training. This results in early detection of biases and possible future drifts. Moreover, these bias detection codes have been open-sourced by AWS.
Clarify goes further by informing how feature values contribute to a predicted individual outcome or for the overall model. This could give a clear insight into why a model might give a certain prediction at inference. A clear practice like this is probably why it was named Clarify in the first place!
Analytics, another big offering at PREDICTif, also took stage with multiple sessions regarding migration of large analytics systems to AWS while modernizing throughout the process & improving analytics productivity for over-whelmed data teams. Not to get lost on all these initiatives, there was a particular leadership talk that caught my attention which talked about creating efficiency and advantage in your data lake architectures, so your data lake catalyzes your decision making process and not idle into yet another data repository.
One big business philosophy that I take away from the 2 weeks of re:Invent thus far is the realization of the industry that multi-cloud could be the true answer. This, coming from the cloud leader itself, might seem ironic but then again this is why it is called re:Invent.
Here at PREDICTif, we realize technologies are never stagnant. It is the name “re-Invent” that we share our values with. Always expanding, always inventing.