Hi
I work in big tech ML org, there are just too many layers and I only end up know things on a high level (product hybrid archetype) but lacking deep technical skills in core ML areas.
Would picking a opensource project to learn deeper technical skills help in interviews ? or Just moving to core area team is better even when sub-org changes and is kind of promo timeline reset.
Given that you're already in big tech I presume youre only targeting big tech. for most big tech MLE roles it's really just a regular SWE loop + 1-2 rounds of ML focus. So you don't need anything overkill for clearing ML interviews at a big tech level.
I suggest just giving it a shot anyways instead of first trying to get focused experience/doing something as drastic as transitioning teams.
With the caveat that I'm not sure how easy it would be to do this at the senior level. From what I'm seeing getting senior SWE roles at big tech in itself is hard right now in this market so landing SWE MLE without prior core MLE experience might be more challenging. Though definitely possible at the mid-level
I recommend these resources:
For ML depth. https://www.amazon.com/Designing-Machine-Learning-Systems-Production-Ready/dp/1098107969 (I know the book is about system design but I think it does cover a lot of fundamentals needed for depth)
Here is a cliff notes for the above: https://github.com/serodriguez68/designing-ml-systems-summary
For ML system design I suggest byte byte go's ML system design. I've gone through this and it was fantastic. https://bytebytego.com/courses/machine-learning-system-design-interview/visual-search-system
I assume you're a machine learning engineer in your Big Tech ML org? If you feel like you're not getting the depth you need, you should switch teams. Open-source can help, but hands-on business experience, especially at a top company, reigns supreme. Availability of options is one of the biggest benefits of working at a Big Tech company.
Another thing you can do is just apply to a bunch of ML roles and see what happens. Maybe your lack of knowledge isn't as bad as you think. I have seen so many engineers underestimate themselves π