Gluttony of great open ML tools too hard for enterprise to use
Seems like every week we hear about yet another new open source Machine/Deep Learning library or Analytical Framework.
Talking to people at Strata this week only confirmed for me that in the midst of what can only be described as virtual gluttony of open-source software, there is massive number of organizations who find it increasingly harder to implement these technologies. Even the task of identifying the right solution can overwhelm many, and result in a tailspin of endless use-case/feature comparison.
In the meantime, the gap between access to open and capable data science software, and the necessary know-how to actually use it, is growing. This trend is only accelerated by constant innovation in the open-source community. How many ML libraries do we have now?
While we wait on consolidation, there is a massive opportunity for streamlined and opinionated data analytics platform to emerge. Platform which focuses more on usage patterns, and less on software package delivery.
To some degree, Cloud Service Providers have an advantage here due to their ability to on-demand provision the necessary capabilities backed by their own elastics resource pools. Regardless however how such platform is delivered, the provider has a window of opportunity right now to earn those net-new analytical workloads by allowing developers to focus on building their differentiation.
Over time, if that provider continues to run these workloads cheaper, faster and with all the necessary level of control, they will have created themselves an organic and loyal customer base.