Transitioning from Software Engineering to Machine Learning Engineering

Learn about the key mindset differences between Machine Learning Engineering (MLE) and Software Engineering (SWE). Mustafa Ispir shares his experience transitioning from Software Engineer to Machine Learning Engineer and the importance of understanding the different complexities within each subset. Find out what a typical workday looks like for each role and how they differ, from planning to defining success.

Resources:

Chapters:
0:00 - Intro
1:41 - Implementation vs experimentation
2:41 - Coding vs Analysis
3:27 - Modular interface vs complex dependencies
4:38 - Functionality vs quality
6:01 - Unit Test vs evaluation
7:21 - Complexity vs complicated
8:27 - Recap

#TensorFlow #softwareengineers #ml

Products mentioned: TensorFlow - General, TensorFlow - TensorFlow Core

[TensorFlow]

Are you considering making the shift from SWE to MLE? Have any questions? Drop them in the comments below!

[skoppisetti]

I made the transition just a few years and I can relate to everything you said. The biggest adjustment I had to make was in the delayed gratification of your efforts. You will never know if all the countless hours you spend on experimenting with an idea will ever come to fruition in contrast to my software eng. days I knew what I was aiming for and exactly how to get there.
I lead an ML team today and I think I try to use the advice you gave. My mantra is "Try and fail rather than not trying at all". The number of experiments is a measure of performance rather than the number of successes.

[boxes]

I never expected a video to suddenly solve my dilemma of SWE vs MLE that has been going on for a few days now. Superb!

[wayne]

As a professional software engineer working on personal machine learning projects, this just gave me a huge "ah ha" moment. Even on small projects, your mindset has to shift. 🙏🙏

[dartneer]

This is probably the best video and should be the first video every SWE watches before diving into the world of ML. Thank you. Thank you! x3

[andriipcreate]

Im a data engineer, and I also study ML in the cloud. If it's needed I can proceed in this domain, it is very exciting.

[pevprague]

This video is great. The biggest issue with ML is that there is basically no modularity. As soon as the output of one model is changed (dimensionality, non-linear output values transformation, ...) and the output is an input for another model/s, the entire cascade of models needs to be retrained. (which is time-consuming and brings a lot of trouble when re-adjusting dependent models).

[DK-oxze]

Great insights. I am an experienced software engineer and learning machine learning currently. I want to transition to ML full-time, but I’m wondering how the levelling looks like after that? Will I have to start from entry level ML job? How did it look like for you?

[ahsanmohammed]

I have two sons in the University of Toronto.
One in computer science
One in computer engineering
Both intend to do a two year masters

What should they do in AI in their masters, other than ethics and cybersecurity?

Provide five areas to choose from.

Thank you.

[mrfancytanks]

What does one need to do in order to land their first role as an MLE? (Assuming one has studied machine learning and is a SWE)

[wryltxw]

Great content. Like to see more content like this.

[pedramezzati]

great, I am more clear about it, but still I dont know should I move to MLE or I should continue working as SWE ?

[julbak]

Any tips for switching from classic computer vision to ML-DL? I've been reading the State of the Art papers and other videos.

[EquinoXReZ]

Well now I don’t want to become a ml engineer 😂