Data Scientist as a Service

One takeaway from AI NextCon 2018 is about abstraction layers on AI (ML and Deep learning).

Yesterday at AI NextCon 2018 Seattle, we were presented “Operating system for AI” by by “Algorithmia”. The presentation brought a nuance on allowing data scientists to sell their AI algorithms like APP developers do today on Mobile App store.

The system allows Data Scientists to upload their algorithm in a chosen programming language and takes care of algorithm monetization, discovery and deployment. Algorithm store allows seamless purchasing experience for enthusiastic buyers using API to interact with algorithm and their data. There are some nitty gritty details, on how Alg. Seller (Data Scientist) enforce the data rules/limitations and structural conformity. Setting those details aside, It is a nice framework to bring data scientist and Businesses together in building AI infused applications to the customers.

After the presentation, I spoke to Algorithmia on the side, on how the narrative may not fit operating system for AI. Operating system meant to me of taking care of things under the hoods at deeper scale managing the complexity, which is not what Algorithmia presented. It appeared for me, Algorithmia is providing an abstraction for their customers in purchasing Data scientist services, “Data scientist as a service”. During the discussion , I asked the presenter if title “Data scientist as a service” fits better as a title than OS for AI. Presenter loved the concept and the name , said he would use that for next presentation.

Any new technology takes its hype curve and takes years to get leveraged by businesses. I see beginning of Enlightenment of AI technology hype curve , as abstractions are built on top of technology to make it useful. Transfer learning technology is a major step towards that. AI model builders don’t have to have gigantic dataset to build models, they can transfer learning from existing models.

Google, leader in AI technolgy has multiple abstractions for AI, like APIs , Cloud ML engine now. Google presented AutoML in the AI NextCon 2018 conference, an abstraction that leveraging Transfer learning technology. Customers can leverage models that are already trained on huge datasets, but can use smaller custom data specific to their use case.

It only is a matter of time, general data scientist services not just specialized algorithms are monetized at per API call basis. “Data Scientist as a service” would take next step in customer adoption of the AI and providing their customers AI infused services/applications.