Machine Learning Jobs are Growing on Trees

The consensus of several discussions at ICML is that the number of jobs for people knowing machine learning well substantially exceeds supply. This is my experience as well. Demand comes from many places, but I’ve seen particularly strong demand from trading companies and internet startups.

Like all interest bursts, this one will probably pass because of economic recession or other distractions. Nevertheless, the general outlook for machine learning in business seems to be good. Machine learning is all about optimization when there is uncertainty and lots of data. The quantity of data available is growing quickly as computer-run processes and sensors become more common, and the quality of the data is dropping since there is little editorial control in it’s collection. Machine Learning is a difficult subject to master (*), so those who do should remain in demand over the long term.

(*) In fact, it would be reasonable to claim that no one has mastered it—there are just some people who know a bit more than others.

28 Replies to “Machine Learning Jobs are Growing on Trees”

  1. And of course the main reason is that, unlike most of other AI sub-domains which have promised a lot for many years but didn’t produce money, machine learning is actually bringing big bucks for those companies that use it extensively.
    As a counter example, I remember this article found via that explains that the word robotics actually makes venture capitalists NOT invest in those companies that use it.

  2. In my experience, companies are using a lot of statisticians. Machine learning is another way to create the same value but for cheaper (since less people are involved).

    One caveat though: easy to master machine learning techniques can produce a big outcome. More evolved technique does not increase so much the added value.

  3. It would be really interesting if someone with expertize in this domain could make a more detailed profile of such enterprises and maybe give some examples.

  4. Funny you mention this. I was talking about this with an undergrad recently who was trying to decide which direction in CS he should focus on. He was trying to find out whether to into Computer Security / Cryptography or Machine Learning / AI. One of his considerations were the job prospects. So he told me about a little search he made on the usual job sites (monster etc.) and how he came up with many security related jobs but none for machine learning and a hand full for data mining. The data mining jobs sounded more like Business Major things, i.e. Customer Relationship Management with SAS etc.

    Is there a job-website somewhere for all the Machine Learning jobs? Besides the obvious Companies like Google, IBM and Amazon I didn’t quite know where else to point him to.

  5. I work at Double Click (or rather almost Google) and yes we use various Machine Learning and Statistics methods to enhance ad serving for our clients. My group is the only one that does ML and is growing faster than any other group in the company (in terms of revenue at least). We are also hiring pretty strongly and having a hard time filling the positions.

    After years of questioning by friends “But Math…what are you going to do with that?” I feel as if it is finally clear what us ML, AI applied math and stats folks can do… sort impossibly large mounds of data, make some sense of it, and turn it into a functioning product!

  6. Can someone recommend some good Machine Learning resources (books, sites, etc.)?

  7. I was wondering if any of you could tell me what would be the best ways to prepare for a career in ML. Is there a particular degree level that’s more prevalent (PhD vs. MS) and in which subjects (is it mostly Computer Science or Statistics… or something else)? Also, could you tell me what kinds of work experience or software experience would be best to prepare me for a career in ML? Are there things I could study outside of a degree program that would give me the knowledge I need? I’ve always enjoyed Statistics & Probability and now I’m a computer programmer, but ML looks incredibly interesting to me. Thanks for your ideas.

  8. For a sense of businesses that hire, check out, in particular their newsletter. The boundaries between knowledge discovery/data mining and general machine learning are fuzzy; one could argue they’re mainly that the KDD world has more a) emphasis on application-driven problems, and b) presence of the business world.

  9. [1] For quants, focuses on VaR; but the site is generic enough, and interesting enough, that it deserves attention.

    Barry Schacter has had the site up a number of years now. He has a weekly email newsletter, with current openings listed at the end. Job-specific; the first link below, then the general site link, where you can sign up for the weekly mailing:

    I typed “machine-learning” into the search line and got four hits; but that’s not the limit. The first link above links to two interesting articles, job-wise, along the line JL posted:

    [2] A totally different thing – but useful – a favorite specialized and useful site: For background; Google Scholar is now returning patent hits; and there’s a separate Google Patent search possible – but how do you get from tif patent pages with AlternaTif giving a page-by-page review, to a pdf download?

    The easiest thing I have found, if you are researching patent literature, for that kind of downloading:

    Simply cut the patent number off Google, and paste it into the pat2pdf search line. (Or am I missing something, where you can now simply get a patent pdf download, from Google?)

  10. I would highly recommend the newly published book “Pattern Recognition and Machine Learning” by Christopher Bishop. It is fairly mathematical, but the author is an excellent writer and the book is extremely up-to-date.


  11. I think that if you’re only interested in Machine Learning, then a Bachelor’s degree in Computer Science will not help you that much. In a typical computer science degree curriculum, you will spend a lot of time learning to write programs, build compilers and operating systems, design network protocols… these will not add much to your understanding of machine learning. A degree in Statistics would be better since a lot of practical machine learning techniques originate from applied probability and statistics. However, if you want to keep your options open, then a Computer Science degree is better since you will touch upon so many fields and will have more career choices to choose from.

  12. I’ve seen the job listings and recent surveys over at KDnuggets, but I wonder what the wider distribution of such jobs is by geography and job type. I suppose one issue is the type of job, too. My impression of most pharmaceutical statistical work is that the analyst follows cookbook rules (boring!).

  13. I was just thinking if someone could help me decide on taking Machine learning as my focus in my graduation. Currently I hold only a Bachelors in computer science. I’ve always loved statistics, probability and machine learning (never went deep into ML). But, then I moved to a job which involves large scale optimization,TSP and stuff like that. Now my original interests are surfacing again and I want to pursue PhD in Machine learning.
    However, I would like to join research industry rather than joining academicia. I’ve read elsewhere in posts that currently machine learning jobs are less compared to data-mining and statistics. But, completing getting my PhD will at-least require 5 yrs.
    Could someone throw some light on the expected job markets for ML researchers after 5-6 yrs? Whether progress in this field has been really fast or slow? It would be great.

  14. It comes down to whether you want to be a Mathematician or a Software Engineer. Most computer science graduates are confused about who they are and what to do with their knowledge. Many switch to other careers after they spent years studying “Computer Science”. The reason for the confusion is that what is currently being taught at CS departments is not ‘Science’, because ‘Computer Science’ does not exist, it is simply “Software Engineering”. Science, such as Physics and Chemistry, is the study and discovery of physical principles that exist in nature. You cannot discover natural physical principles in a computer which is a man-made artifact. Unfortunately, because of the confusion, most computer science graduates lack competence in both science and engineering.

    So at this point, you are better off asking yourself whether you want to be a Mathematician or a Software Engineer. If your interest is in the study and discovery of mathematical principles that allow inference from data, then you should work to become a competent Mathematician or Statistician. If on the other hand, you’re interested in applying the principles discovered by Mathematicians to solve real world problems such as recognizing a face or a voice, then you should work to become a competent software engineer. By being competent, there will always be demand for your skills.

  15. I would also throw applied or computational mathematics into the mix–I’m sure I’m biased because this is my background. Make sure it has a computational bent so you don’t wallow in the world of Ring Theory!

  16. Surely engineering implies building something, CS has a strong maths component and is thus more theoretical than straight Software Engineering, for which courses are also available.

    Of course in CS you will have to cover a lot of ground (including SE) so the focus on ML could be quite brief.

  17. I think computer scientists fall into more than two categories. You have the applied mathematicians, and the engineers of course, and the pure programmers which I’d claim are separate from engineers. But you also have theoretical mathematicians. While ML (at least the sort you get paid to do in industry) tends to focus on statistics and calculations, there are other topics which focus on algorithms, optomizaitons, logic, etc… For example: “Given x information, devise an algorithm to solve problem y. Can you prove that your solution is always correct, and if not, can you formally characterize its errors? Can you prove that there is no more efficient solution to the problem, or, otherwise, can you improve the algorithm?”

    This sort of thing shows up in compilers, systems, distributed systems, database theory, theoretical ML, etc… and is kinda fun :-).

  18. I recently finished my Msc in AI/ML from a good department in the UK. I was surprised to see how much I struggle to find jobs after my degree (I have a good BS Computer Science + experience). Note that I have been looking outside the IT industry such as in finance and trading, applying mainly for graduate modelling jobs. I find that both recruiters and many employers have a distorted view of what AI/ML is, not realizing the high statistical content of machine learning. I have a PhD offer on the table for next year but the my current job problems are causing some doubts. If doing a lengthy PhD I would like to be sure that my job prospects would include other industries than IT. I am even considering taking another Msc, now in pure statistics to be able to catch those modelling jobs I am most interested in, although somewhat frustrating. So should I do another Msc. or would the ML PhD make me as much or more qualified for the same modelling jobs.

  19. After doing a Msc ML I can vouch that jobs are very few and far between. Wish I went with a Stats instead.

    1. Just to clarify, you did an MSC in machine learning and the jobs are very few and far between?

  20. As a UK university we have always wanted to introduce a MSc in Machine learning, or AI. However when we considered the market demand and employability upon graduation, we had to change our mind. It’s a great surprise to me, because people should be using AI to automate processes in this day and age.

  21. really can you post some companies website links who are hiring…..

Comments are closed.