labels: career_advice, reddit, mentoring, early_adoption, autobiographical
This is more of a career question, so hopefully, it's the right place to post it, but let me know if it's not!
Senior ML Eng and 5+ years of experience. I've recently been feeling very stuck in my career progression. Throughout my career (and especially recently) I've never been able to work on any incredibly challenging machine learning solutions. I get stuck working on problems that have fairly mature solutions (regression, classification, recommender systems, etc.) in orgs that have little appetite for more complex solutions. I have the desire to work on more deep learning projects, but in my experience, this work usually seems reserved for those with Ph.D.'s or a large amount of prior experience in the field. I understand deep learning fairly well, but don't have a ton of professional experience working on it.
Has anyone had any experience with getting an opportunity to work on deep learning software without having an advanced degree?
You're asking at the perfect time. The ML space is experiencing a lot of flux right now, which means there's a lot of opportunity for people like yourself to position themselves as short-list experts of emerging technologies and techniques simply by virtue of being an early adopter. It doesn't matter if you don't have a PhD if you're one of five people in the world who has ever played with X new ground breaking technique, and being in a small group like that also creates a lot more potential for valuable and even novel contributions. I do have an MS in statistics, but the vast, VAST majority of value I've contributed to the world has just been a consequence of me being curious enough to keep up with research and making it easier for people who aren't to see and play with new developments.
Concrete example of the kind of thing I'm talking about: AnimateDiff is a recently described technique for text-to-video animation. The authors kindly shared their code and pre-trained models, but they did it in a way that isn't particularly "artist friendly", which had the consequence that a lot of people who were interested in this work didn't have a chance to play with it. So I spent a few hours refactoring their work to slap a simple UX on top of it, and now that notebook got so much attention I'm already regularly explaining to grateful people that all I did was massaged what the authors had already provided.
There is an overwhelming flood of amazing work happening right now, but our FOMO reflex has a tendency to draw our attention to just those things that it seems like everyone is talking about or playing with. Researchers are people too, and people only have two hands. There is a lot of great stuff out there that's not getting the attention it deserves simply because the people who published it didn't make it easy for others to play with it. Keep an ear to the rail for interesting looking projects, and if you notice that no one seems to be playing with something that looked really promising, that might be an invitation to use your engineering skills to help get the word out.