Clojurians Log v2

I’m just getting back into interviewing but, so far I haven’t found that to be the case.

I will precise: "developer has almost zero chance" This sentence is my conclusion, because market demands Python, PyTorch, TensorFlow, Pandas etc. which are on the first place Data Scientist experience to teach AI, not software developer implementing AI. But from what I heard all corporations demand this on first place anyway.

It was literally said there is no way to get such job without Python, because everyone demand it.

I'm skeptical. "almost zero chance" would be true only if: • the recruiter side is perfectly competent in evaluating candidates • no one in the recruiting process (both sides) ever attempts to oversell themselves or bullshit each other • the job description (either intended or imagined or written or all of these) perfectly matches the actual work • many others discrepancies and dysfunctions never happen

1. Most data scientists are not very advanced when it comes to PyTorch (et al). They have experience with short-ish scripts, and using Python by calling ready-made procedures. Regarding that, you can catch up quickly IMO. 2. OTOH, most developers are even less experienced when it comes to ML, AI, statistics, and math. That's what will be tricky for you. 3. OTOOH, a cynical me will say that even most data scientists don't have a good knowledge of the foundations, so you may still have some chance. 4. OTOOOH, getting the foundations right is not as easy as it seems, so....

But yes, they'll probably look for the right keywords at least, so you better have those with you.

For what it's worth, and bouncing off blueberry's point about the technical erudition of the average data scientist, it should be a not insurmountable task to take something from a white paper and use it to train a model using lib-python-clj or Neanderthal, or any other ML api available to Clojure to demonstrate your ability to build a non-trivial training pipeline in Clojure, with or without Python

I understand rationale, but all this doesn't matter if companies demand this experience to hire you. My question is, if this is true and most of companies (if not all) demand this experience for AI implementation architect. So far from what I saw it is true. At least this is my impression in EU, but this is observation on a small number of samples. On the other hand HR saying the same. I am trying to understand companies demand in AI.

@kwladyka You can use all that python tooling you mentioned from inside clojure. You can also carve yourself a niche without python, given the track record of clojure shops doing AI/ML without Python. You probably would have less luck avoiding those tools applying to a well established ML/data science team than you would as a freelance contractor on a greenfield, but Python is not the be all and end all of DS/AI by any stretch. And like others have already mentioned, it sounds like you are being a bit pessimistic to insist on your conclusion while focusing only on your own anecdotal evidence after you have asked for other people's input. If you want to find a niche in the market that isn't dominated by Python, then you'd probably be better served either seeking out people who are doing data science/AI in Clojure or the wider field who aren't using Python than fixating on the requirements of the average AI role.

There is misunderstanding here. My question is not technical or how I can code better AI or find a niche or who can has better experience in tools or... The question is about experience required by companies for this role and who they really hire. Not by a few niche companies, but general trend in US and EU. The question is only about hiring process. Sure I am suggested by HR about hiring process. Shouldn't I? This is why I am asking here to get another opinion. I got multiple answers, but I don't feel any answered my question. Perhaps I made question wrong or nobody here really know.

@kwladyka >general trends Thats exactly my point. If you fixate on the tools that are in majority demand, you will never find an opportunity to use any other tools. If you want to take the impossibility of getting hired to build AI pipelines and train models outside the most commonly used Python libs at face value, then it becomes something of a self fulfilling prophecy and no amount of asking other people for input will help you break free from that mould. Plenty of companies exist that do not use Python to train and implement their AI systems. Ergo, it is not functionally impossible, as you say, to find an AI job that does not require you to use the Python toolchain. How far do you think I would have gotten 10 years ago if all I worried about was the fact that a significant portion of the SWE market is focused on python and JS instead of focusing myself on Clojure and aggresively seeking out Clojure roles? Not very far, I feel like. I would assume blueberry is worth listening to on this, since they are a maintainer for AI training tooling in Clojure and might know people who are using it at work.