|
I've been studying ML on my own for a couple of months now and really enjoying it. It's pretty impressive what you can do with some deceptively simple algorithms. I've been enjoying it enough that I'm thinking of migrating from web development, which I've been doing for over 10 years, to a more data-centric/ML kind of role. I hear that there's a lot of demand for these skills now but a perusal of various job boards seems to indicate that this is sort of a niche field. I don't want to go down this road if I'm going to be competing with people with Ph.D.s in the field for every application. Can anyone offer some insights into the size and health of the job market for these skills? Is someone without formal qualifications out of luck? |
|
I've been evaluating the job market for NLP and ML for the past nine months. The demand is huge, and only looks like it is going to get bigger. Demand certainly outpaces the number of job applicants, as evidenced by the job openings that have been open for a while. This is true even if you have industry experience or an MA instead of a Ph.D. The main difficulty is that a lot of companies want their NLP + ML expertise in-house, and they assume this means they need Ph.D.s to work long-term, full-time, on-site. But many NLP + ML Ph.D.s don't necessarily want this style of arrangement. The problem is that many companies that need ML + NLP don't need Ph.D. level breadth of knowledge. So a lot of positions at good companies remain open, because the Ph.D.s don't want to work on a narrow problem with little research, little exploration, and little leeway. What actually makes more sense is to hire very bright programmers in-house, and have a business-savvy Ph.D. advise the programmers about the specific task that the companies works on. The advisor steers the programmer away from pitfalls and dead-ends, and soon the programmer learns about the task of interest. ML and NLP aren't actually that hard once you know what you're doing, it's just that there's a huge variety of techniques to learn if you want enough breadth to tackle arbitrary problems. But most companies only have very specific problems. That's exactly the kind of arrangement I'd love to find. If I could work with a good mentor for a year or two to help guide me it would definitely accelerate my learning process. I've implemented several ML algorithms on my own already but someone with some insight into the underlying theory would be a big help.
(Sep 23 '10 at 10:27)
Miles Egan
|
|
Generally the demand for Machine Learning and Data Analysis is larger than the supply of PhDs in those fields. There are plenty of opportunities to provide value. As long as you can do that, it shouldn't matter what anybody else is doing. http://www.kdnuggets.com/ is a good resource for what the market is like. The website www.kdnuggets.com is really awesome! Thank you,zaxtax
(Mar 05 '14 at 00:30)
Miao
|
|
For every commerical ML job that demands a PhD there are probably at least 2 or 3 jobs dealing with the infrastructure to support these ML efforts. This includes robust data collection, implementing algorithms to deployable quality, setting up and managing computing clusters, and so on. With your programming experience you could probably move into these roles fairly easily. Searching for jobs that mention "Hadoop" or "Map/reduce" should turn up lots of jobs that don't require a PhD. That's an excellent point. I'm pretty confident in my software engineering skills so a role like that would probably be a good starting point.
(Sep 23 '10 at 10:28)
Miles Egan
|
|
How do you feel about pursuing a masters degree with a focus on data analysis (data mining, machine learning, statistics, operations research, ...)? My sense is that there are many interesting job opportunities for people who are comfortable working with data and who can make use of algorithms from the fields listed above. Some require Ph.D.s, but many do not. Even a masters may not be required, provided a job applicant has demonstrable experience in these skills. But put yourself in the employer's shoes for a moment. How do you figure out if a candidate has the analysis skills for the job? Do you filter based on having a masters with an appropriate focus? My suspicion is yes for the kind of job you describe: most people with only bachelor degrees will not have had a lot of experience with data crunching or the algorithms to do data crunching. If this area interests you, the masters degree has the advantage of systematically exposing you to a broad range of techniques . . . perhaps more than you could easily study on your own. It will also likely help you develop contacts for finding the jobs best aligned to your interests. A master's is definitely a possibility. The other option is building something demonstrating some solid knowledge of the principles. I'm still not sure which would be a better investment of time. I'm planning on spending the next three months or so studying on my own so maybe I'll be in a better position to evaluate a master's degree afterwards.
(Sep 22 '10 at 20:25)
Miles Egan
|
+1 Thanks for asking this question. I am in a similar situation in that I have been a web developer and would like to transition to NLP/Machine Learning/Computational Linguistics. I am currently in the midst of PhD study in this area, but my day job is completely outside the field and it is so frustrating.