What is MLOps and how to get started? | MLOps series

❤️ Become The AI Epiphany Patreon ❤️

👨‍👩‍👧‍👦 Join our Discord community 👨‍👩‍👧‍👦

Disclaimer: this channel is separate from my work at Google DeepMind and is my attempt to educate the broader community of future ML engineers, researchers, and builders.

With this video I'm kicking off a brand new video series on MLOps! Everything you need to know to build an ML-powered app end-to-end with a high level of automation.

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
The 4 courses I mentioned:

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

⌚️ Timetable:
00:00 Intro: great resources for learning MLOps
05:25 What is MLOps? (leave feedback - what do you want to see?)
13:30 Walk-through of MadeWithML lectures
33:08 FSDL (full stack deep learning) course
35:39 Deploying ML models in production Coursera course
38:30 Outro

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
💰 BECOME A PATREON OF THE AI EPIPHANY ❤️

If these videos, GitHub projects, and blogs help you,
consider helping me out by supporting me on Patreon!

Huge thank you to these AI Epiphany patreons:
Eli Mahler
Petar Veličković

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

#mlops #ml #apps

[bitcoingrinding]

I would really like to see you doing a end to end project covering all those topics :)

[anandpawara]

I have been waiting for this series for a long time... thanks a lot

[aleksabisercic]

Love it! Looking forward to the series.

[SnipeSniperNEW]

I just had an MLOps interview and I did horrible. Time to actually learn MLOps the right way!

[SinanAkkoyun]

YES, the only youtuber actually diving deep in, we need you!

[sabaokangan]

Thank you so much for sharing this with us on YouTube ❤️‍🔥

[lukasugar]

This looks like the most useful series you've done so far!

[musa_b]

I'm glad to see a video series on MLOps! I love how it utilizes AWS, but for those without access, there are alternative solutions like using Docker or Kubernetes on your personal computer. This way, you can gain deeper insights into the infrastructure and tailor it to your individual requirements. Keep up the informative content!

[agr__]

fantastic idea. I will follow this with great interest

[tripathi]

I am super duper excited for this series. Please cover everything that you outlined today, already looking farword for next episode.

[gattabat]

Just came back to watch this again - incredible

[EkShunya]

thanks you for all the effort you put into these lectures, it is a great learning resource for many.

[muhammadhammadkhan]

Super excited for this series looking forward to learn everything about MLOps 🎉

[lukasugar]

There're so many MLOps topics to choose from... I'd say that experiment tracking and dataset versioning are the topics I'd be most interested in :)

[Chimecho-delta]

Looking forward to this! (As a software engineer that just follows AI research, particularly alignment)
I think my preference would go to a more practical series with a concrete example using technologies and actually deploying something. (big fan of aws lambda haha)
Like you mentioned this being MLOps I'm not looking for info on tuning/developing a model.
However, setting up a training environment in the cloud/orchestrating the training on multiple TPUs/GPUs would also be interesting to know more about. (If you see this as MLOps that is)

[stevengeorge]

Awesome video! I've come across some of the links you cited but never really knew where to start so looking forward to your future videos!

Some things which I would find useful in future videos are:

1. How to take a trained model and deploy it on the backend. I can train models but I'm often unsure of the best practices for deploying the model e.g. using Docker and where to store the model

2. How to develop better front-end applications. I've used Streamlit a bit but would be great to know how to develop nicer interfaces and features like the ones on your Huberman app

3. Options for cloud GPUs. Colab is helpful for rough notebooks but not great when it comes to versioning and working with multiple Python files. I prefer to SSH into a GPU instance with VSCode (don't want to purchase my own GPU) and I currently use GCP for this. However I'm not sure if this is the cheapest/best option out there and the number of configurations when setting up an instance is quite overwhelming. Would be great to understand this a bit better

[joseortiz_io]

Really excited about this series! Definitely interested in all of these topics. Particularly interested in Data versioning, containerizing and serving and monitoring to trigger retrain. Okay maybe all of it hahaha 🤖

[MLOps]

nice one! this is great thanks for putting it together

[chiragsharma]

Super excited for the series. Weights & Baises has also offered a free MLOps course last year in December in case one wants to check out. It's completely free now as well. I just discovered about it yesterday.
Aleksa please only try to touch upon the model training only a bit so that we can see how a project is implemented.

[arda]

In the next video, you could go over the steps you took in Andrew Huberman's Transcripts project. A detailed explanation will be very nice.