ModelOps History

To know the truth of history is to realize its ultimate myth and its inevitable ambiguity. -Roy P. Basler

Did you know! It took the automotive industry six years or more to design a new car in the 1990s.

So, it all started back in those days when we had this really cool team always producing

bug free code - just kidding :)

Well, in my early years of development, we used to develop a product feature and, point to point, write an operational document. - by the way, this document did not have any details about the code, rather it only illustrated how to execute the code.

We shared the same document with Operation Team - I call them brick wall - to try repeatedly to do something with no success. These guys are so tied up with the operational document that they had nothing to say each morning except "It felt like I was hitting my head against a brick wall because I had no support from my boss". - lol

Both developers and operations departments had separate objectives, separate department leadership, separate key performance indicators by which they were judged, and often worked on separate floors or even separate buildings.?

Nevermind, Soner or later around 2008, we adapted Agile model, leaving behind Waterfall approach - Things were really getting exciting in terms of deliverables - business started getting small packets of continual value. But, it increased operational challenges with every release. We enhanced the same code to a degree - too many documents to be updated, execution & deployment process changed - conflict between both department leadership was unavoidable.

Yes! that when DevOps was born from the collaboration of developers and operations leaders getting together to express their ideas and concerns about the industry and how to best get work accomplished. DevOps touches every phase of the development and operations life-cycle, from planning and building to monitoring and iterating.

Embracing Agile!

It was around 2010 that I end up working with RBS (Royal Bank of Scotland) Edinburgh handing almost 70 plus data servers (along with 4 other colleagues) that comprise data technologies likewise Informatica Power center, Power exchange, Golden gate, Oracle, DB2,.. etc - and next big thing was talk of the town

1. Big Data

I literally happened to live DataOps at least 4 years before it was first introduced by Lenny Liebmann, a contributing editor for InformationWeek, in an article titled DATAOPS: WHY BIG DATA INFRASTRUCTURE MATTERS

Any ways, as Data application grew on the same lines of DevOps, the term “DataOps” was born. Ta-da...

This new emerging discipline made up of data engineers (ETL Developers) and AI/BI Engineers. The principal aim of DataOps was to improve the company’s IT delivery outcome by bringing data consumers and suppliers closer.

This is hilarious - how Business sought help from "IT" to process their data and as time passed by its "IT" that owns the Data, and today business or knowledge workers depends on "IT" to make their data available. Seriously!

Whilst this all was happening, we had a second story running on the other side of the road.

• 2011 Apple releases Siri, an AI-powered virtual assistant, through its iOS operating system
• 2014 Google makes the first self-driving car to pass a state driving test.

Whilst they carefully crafted ML model in a controlled development environment, data comes from that continuous degenerated sources known as “the real world.” It never stops changing, and you can’t control?how?it will change.

Datarobot claims to be AI platform that democratizes data science and automates the end-to-end ML at scale.

Getting a model to work great in a messy notebook is not enough.

This fundamental disconnect causes several important challenges that need to be solved by anyone trying to put an ML model in production successfully.

That brings us to?MLOps. It was born at the intersection of?DataOps,?and?Machine Learning. A typical MLOps software stack might span data sources and the datasets created from them, as well as a repository of AI models tagged with their histories and attributes.

Add monitoring to MLOps turns into ModelOps !

Multiple and silo team develop MLOps across the enterprise using different tools and processes, what more unless monitored properly, ML models degrade and deliver unreliable decisions that can put business at risk and result in lost revenue.

ModelOp helps to characterize these challenges. Without ModelOps capability, there’s no owner on the operational side.

This is a new and exciting discipline, with tools and practices that are likely to keep evolving quickly.

ModelOps incorporates MLOps, which is managing ML models throughout their life cycle at an enterprise scale. ModelOps is considered as a superset of MLOps.

Curious to find out what is AIOps! post your questions in the comments.