As you may have heard, models are a tiny part of any typical ML-powered application.
There is nothing that stresses that as this picture:
Source: Hidden Technical Debt in Machine Learning Systems,
https://t.co/JDyAr1s3kc
There are lots of critical processes that are involved in MLOps such as:
- Data processes: collection, labeling, exploration, preprocessing
- Modeling processes: building, training, evaluation, testing
- Production processes - Serving, monitoring, and maintaining models
MLOps is a new topic for almost anyone. Maintaining models for a prolonged period of time is difficult.
Models are very prone to change. They drift over time. The world (that sources the data) changes, and so data change too.
MLOps is a huge topic. All I wanted was to have a reasonable understanding of it.
Here are 3 resources that I used:
- Machine Learning Engineering book by
@burkov - MLOps Specialization by
@DeepLearningAI_ - Introducing MLOps book Oreilly
Here are links for those resources:
- ML Engineering book:
https://t.co/L5trxHGAw1 - Introducing MLOps:
https://t.co/de4vxdzA5P - MLOps specialization:
https://t.co/46fhFSyEno
I also wrote a couple of blog posts as I was learning it. You can find the blogposts on Medium
https://t.co/DFp6LwqxRV
If you would like to get started with MLOps, I recommend you take MLOps specialization along with one of those books, preferably ML Engineering book.
Also,
@MadeWithML by
@GokuMohandas contains many hands-on resources for building and productionizing machine learning models.
I can't recommend it enough too!
https://t.co/WjYQpeXcTX
If you have mainly been building models, learning MLOps might be the next good step for you. It's a useful skill to have!
Thanks for reading!
If you would like to see more machine learning content and useful resources, follow me at
@Jeande_d. You can also share the thread with others if you found it helpful. Sharing is caring :)