Hello everyone I am so excited to share with my readers on an important topic MLOps.
Probably you would have heard many talking and writing about this. I am here to give a very simple explanation and my personal opinion.
So let’s start it. I never imagined when I started my career I would be in a place to start a practice on MLOps.
You might wonder what lead me to it and what did I do to get into this. Continue to read…
Skip below if you don’t want to be inspired or bored with my history
I have started my career on Ruby on Rails and later moved to Asp.net mvc, yes you heard it right I was a full stack developer. I like to be creative and still like to produce contents very creative. This interest in me lead to learn MEAN stack. Everyone in my org thought that he is going to get into UI/UX but there came a shock I started to learn about AI/ML.
So what really motivated me?
It was Andrew NG course on “AI for Everyone “ it was so simple and helped me understand that AI could help society and people.
In that single course it gave me so much perspective about AI. Highly recommend for starters. Then I made internal move to work in Intelligent Automation. I love to read technology and started to explore more on Automation basics. It gave me a very different view and experience. I understood that any automation is a replica of manual steps in a very seamless way.
I got an opportunity to take part in hackathon and I presented along with my wonderful colleagues on “Sentiment Analysis”. I was fortunate to move into as Finalist among top 7. Presented to CTO and Chief Architect. (Wow what a feeling that would be). I didn’t win but I won something else which changed my career totally.
I got the opportunity to build the team on Conversational AI. This confidence perhaps helped build the MLOps practice???? I scaled the team from scratch and build 3 digital voice assistants.
I started to know more deep in AI and then came challenges around deployment & maintaining. I think that is the time when MLOps was emerging slowly. I still remember that I was reading about Uber’s Michael Angelo. That is my first blog to read about ML Lifecycle. I was so confident that I started to talk in my org about MLOps.
Now you could ask me is it possible to even start talking about it??? By this time I had good knowledge on 1, 2 & 3 i.e Engineering, Infrastructure & High level ML knowledge. You need to have good understanding about ML Lifecycle sometimes referred as MLDLC. (Machine Learning Development Lifecycle).
Later a year I moved to a new org where I was tasked to start the practice on MLOps. I had amazing team members who helped me achieve this practice. By the time I was leading my brilliant team members they were exploring on MLOps. I was just observing what they were doing trying to get the current state and picture the future state. This made me understand the gaps they had.
Gaps needed to be identified:
Documentation - This was the first biggest piece. There was no documentation we all came together and started to document. Now it has grown into a large repo of MLOps.
Architecture Artifacts - There was no Architectural representation or artifacts to connect the dots and see the big picture. After many reviews with team and senior leaders we formed a framework.
MLOps Operating Model - I understood there was no operating model or any pre sales deck. We started to create one. (Probably I understood I could be creative again).
Demo - We were working on mvp’s but not solid ones. We started to have solid demos to show our customers.
Partners - The more the people and partners you meet you will know the landscape. I have started to interact with all top ML platform providers cloud and independent.
Many Thought leaders in LinkedIn helped me in many ways, sorry couldn’t post their details but thanks for every thought leaders in this space.
Finally we were very matured even got into Responsible AI practice. Developed a product ( Can’t share more details). The idea was wholly original by me. But I gave this idea to my team members before leaving this short term legacy. 🙂
In between there are client engagement which I am not allowed to share, but it was an enriching experience. I have won 3 accounts on MLOps within a year.
MLOps Maturity Assessment:
Every org were keen on getting into MLOps but major issue was they didn’t had clear roadmap, that’s where MLOps maturity comes into picture which helps to move from Level 0 to Level 4.
Interested to know about. Made a video make sure to check it out.
MLOps Explained Easy | ML Lifecycle | MLOps Pipeline | Overview | Enterprise.
Thanks for reading my thoughts on MLOps. I have more to share on AI/ML make sure to follow me.