Advice on applying to a PhD after working for ten years

[am_i_too_old_fr_this]

(I posted this in a few different places, so apologies if you've seen it somewhere else)

 

I'm a software engineer looking to go back to school and get a phd in computer science with a focus on machine learning. I have a BS in a non-CS engineering field and an MS in applied math (where I took a lot of CS classes), both from top 5 institutions in their respective fields. After graduating from my MS ten years ago, I've worked at two startups as a software engineer / manager. The first startup got sold to a large company and the second (current) startup has been gaining traction and I'm now at the VP level managing a pretty sizable team.

 

I do enjoy my job, but I don't love management and I've been intellectually curious about machine learning for a while now, doing a lot of self-teaching on nights / weekends. Originally, it was just to keep up with the current research and see if anything was applicable for work, but as I dug deeper I've found a few areas that I'm very interested in doing further research. I've also saved up enough money that I can afford a grad school salary for a few years and still support my family.

 

Unfortunately, outside of my self-study, I've only worked tangentially with machine learning (mostly using existing libraries and services). I had taken some related classes during my MS, but that was ten years ago. I also don't have any research experience in the field, although I did do some undergrad research during my BS -- unfortunately, no publications and in an unrelated field. Plus it was over 10 years ago.

 

I had a few questions I was hoping to get some guidance on:

 

•  I'm looking to apply for Fall 2019 admission, so I have some time between now and the end of the year to strengthen my profile. Assuming I keep my current job (so I only have nights and weekends free), what are the best things I can do at this point? I had a few projects in mind, but would these help my admission chances?

 

•  I think I can get strong recommendations from my managers at the companies I worked for. They're at the CTO level and well respected in industry, but probably not in academia. I think it'll be a bit of a stretch asking my MS professors since it was so long ago, and they probably won't be able to write anything meaningful at this point. A third option I was thinking about is asking the same professors I got recommendations from when I first applied to my MS -- they knew me well at the time and wrote strong recommendations, so hopefully they still have those LoRs around. Which one of these options would look best to the admissions committee?

 

•  Assuming I can do well on my GRE and my grades are decent (3.9+ both BS and MS), where should I be aiming? I understand the lack of research experience is a huge negative. Should I consider applying to any MS programs with the intention of doing some research and getting stronger LoRs, and then reapplying to phds afterwards?

Edited February 25, 2018 by am_i_too_old_fr_this

[TakeruK]

I think the first questions are: Why do you want a PhD? What do you hope to do with it? What are your career goals after this PhD? 

[spamhaus]

@am_i_too_old_fr_this Wow, you could be talking about me with a statement like that :-)

I was in a similar boat. I had been in industry for 10 years and worked as both an engineering lead and an engineering manager. I also realized that being a manager was not for me. I desired the intellectual stimulation of solving challenging problems. One main difference is that I do not have kids, only an extremely understanding and supportive wife.

So here's what I did. I actually quit work for self-study and to dedicate time for working on machine learning projects. I did a number of projects and actually landed an interview at a top machine learning research company in SF. Unfortunately, I did not get the position, which made me realize that maybe I should get a PhD so I can have a stronger machine learning foundation.

I then found a professor online (deeplearning.net) who was looking for researchers to collaborate with. I've conducted research with this professor by email since the summer. This provided me both a research paper that I was able to put on arXiv (with plans to publish in a top conference), and a LoR from a professor that was intimately familiar with my abilities.

While I had contacts at top companies in SF, both a VP of Engineering and CEO of a well known public company, willing to write me a LoR, I did not feel those were appropriate for trying to get into academia, as they would have only known me as an engineering manager. My understanding is admissions committees value people who can speak to your aptitude for research. Rather, I spoke to a professor I've kept in contact with over the years since my MS and asked him to write one of my LoR. I also, received one from a former colleague who had worked her way up to a VP at a different company, but was intimately familiar with my ability to research challenging problems.

In terms of where to aim, well you can see all the places I've applied in my signature line. I applied to 15 of the top programs and I've received admission to one (which I'm very please with), I've been denied by nearly half, and I am still awaiting a response from the other half (though at this point they are likely rejections). The university I was accepted to, the POI I interviewed with specifically called out that I was a non-conventional applicant. It could be that this worked against me some other universities, but this professor was willing to take a shot since we aligned very well in research goals. That is, I have a very specific set of research goals in mind, and I've been interested in that area of research for a long time, which was clear in my application.

Here's how I decided where to apply. I looked at ALL the programs that were ranked in the US on csrankings.org, and evaluated each professor to see who I would be interested in working with. I kept a spreadsheet and ranked all the professors of interest on a scale of 1-4. Interestingly, just by looking at the research and publications coming out of the universities, my interests most closely aligned with the top programs... go figure. I have a suspicion that those programs have enough money, that they have the most freedom to pick research ideas without worrying about funding.

I know this post is a bit scattered, but I hope it helps you in some way. If you have any specific questions I didn't answer I'd be happy to fill in the gaps. Best of luck!

Edited February 25, 2018 by spamhaus

[am_i_too_old_fr_this]

@TakeruK I want to get a PhD so I can do research in machine learning, either in an industrial or academic setting. During the last few months I've been reading through relevant research papers, trying to reproduce results, and learning about all the things I don't know yet. Then, in a depth-first-search kind of way, I try to fill in the gaps the best I can. It's a long process -- I've only gone through a few papers, but I definitely see progress being made! It's been super challenging, but I find the field fascinating and I'm enjoying the experience more than any work I've done in the last ten years, so I'm fairly certain this is the right choice for me.

@spamhaus Thanks so much for sharing your experience. And congrats on the offer! UMass sounds like a great school for NLP. Fortunately, I also have a very supportive wife that's encouraging me all the way (honestly I'm not sure she fully understands what she's getting into yet!). It took some real guts to quit your job to focus on this 100% -- unfortunately, I'm not in a situation where I can do the same, but I'll be doing as much as I can with the time I have. I'm still not super confident in my current abilities to reach out to professors about possible collaboration just yet, but I definitely plan on doing it after a couple more months of self study. I'm also still narrowing down my research interests. Thanks for the pointer about csrankings; I'll definitely take a look. One more question I had is how you sold the story of going from industry back to school. Was your prior work heavily related to machine learning?

[TakeruK]

1 hour ago, am_i_too_old_fr_this said:

@TakeruK I want to get a PhD so I can do research in machine learning, either in an industrial or academic setting. During the last few months I've been reading through relevant research papers, trying to reproduce results, and learning about all the things I don't know yet. Then, in a depth-first-search kind of way, I try to fill in the gaps the best I can. It's a long process -- I've only gone through a few papers, but I definitely see progress being made! It's been super challenging, but I find the field fascinating and I'm enjoying the experience more than any work I've done in the last ten years, so I'm fairly certain this is the right choice for me.

Cool! I am not in the field, but know about the process of grad school admissions. I think @spamhaus has good advice and their experience sounds a lot like the successful non-traditional applicants in my field (I do think there are many things that don't change from field to field). Specifically, the things that stood out to me were:

- Spending time doing research in academia in this area
- Keeping in touch with and building relationships with academics for LORs
- Finding someone who was willing to take a risk on a non-traditional applicant
- Finding that it's the top schools with money that they are willing to "risk" on non-traditional applicants rather than stick with the large number of traditional applicants they would get.

From your questions:

23 hours ago, am_i_too_old_fr_this said:

•  I'm looking to apply for Fall 2019 admission, so I have some time between now and the end of the year to strengthen my profile. Assuming I keep my current job (so I only have nights and weekends free), what are the best things I can do at this point? I had a few projects in mind, but would these help my admission chances?
•  I think I can get strong recommendations from my managers at the companies I worked for. They're at the CTO level and well respected in industry, but probably not in academia. I think it'll be a bit of a stretch asking my MS professors since it was so long ago, and they probably won't be able to write anything meaningful at this point. A third option I was thinking about is asking the same professors I got recommendations from when I first applied to my MS -- they knew me well at the time and wrote strong recommendations, so hopefully they still have those LoRs around. Which one of these options would look best to the admissions committee?
•  Assuming I can do well on my GRE and my grades are decent (3.9+ both BS and MS), where should I be aiming? I understand the lack of research experience is a huge negative. Should I consider applying to any MS programs with the intention of doing some research and getting stronger LoRs, and then reapplying to phds afterwards?

1. I think you should do even more self study and also to try to make contact with researchers and work on a project with them, as @spamhaus did. You may not have to quit your job (yet) but it might take longer than until the end of the year. Ideally, you do enough work with researchers that you can ask them for letters of reference. You want to get your background knowledge and experience to the same level as the typical fresh-out-of-undergrad applicant for your desired programs. 

2. I think you are right that you want to get letters from academics. In addition to getting letters from new connections above, perhaps reach out to your old profs once you have this new project going well. Let them know about your new plans and your progress on the new project. You can get a good letter from one of these profs if you restart your relationship with them and build on that. Finally, I think you might want to consider one non-academic letter too. That background makes you unique so you want to also emphasize that. Ultimately, grad schools want people who can and/or have excelled, even if it is not in the specific area of research. Having a top-notch letter showing that you are really great at what you currently do is helpful to demonstrate that you can also excel in their program.

3. If #1 above doesn't work in getting more contact with academia and research projects, I think a MS is a good way to make these connections happen. It will cost you more money and time but if it gets you to your goal in the end, you may decide it's worth it. 

Finally, for a non-traditional applicant, admissions committees generally want to see a few things: i) demonstrated real passion / interest in the topic, ii) success in the other field/area you're coming from and iii) evidence that you can complete their foundational coursework and produce good research.

For i), I think you need to do more than what you have right now to show the committee that you are serious about this change in field. I think answers #1 and #3 above will help you show that you are committed and passionate about this PhD program. Your SOP essay will be quite important in your application to demonstrate what steps you have taken to make the switch. In addition, clearly articulating why you want to change and what your research interests are will go a long way.

ii) is already mentioned above as answer #2 regarding your letters

iii) means they want more than just a researcher. Getting research experience is important but you also need to get the equivalent foundational training as the applicants with CS undergrad degrees (if applying to CS PhD). Most PhD programs view their students as their generalists first and specialists second (most scholars will view themselves this way too). So, don't neglect the other parts of CS if you are applying to a CS program. Self-study in machine learning will be important but you will want to cover other materials in a CS undergrad program too, if you haven't already! At a bare minimum, if a school really just wants to admit you for research in X, you need to at least convince the admissions committee that you will be able to pass their other grad courses and the qualifying/comp exams. 

[spamhaus]

9 hours ago, am_i_too_old_fr_this said:

@spamhaus Thanks so much for sharing your experience. And congrats on the offer! UMass sounds like a great school for NLP. Fortunately, I also have a very supportive wife that's encouraging me all the way (honestly I'm not sure she fully understands what she's getting into yet!). It took some real guts to quit your job to focus on this 100% -- unfortunately, I'm not in a situation where I can do the same, but I'll be doing as much as I can with the time I have. I'm still not super confident in my current abilities to reach out to professors about possible collaboration just yet, but I definitely plan on doing it after a couple more months of self study. I'm also still narrowing down my research interests. Thanks for the pointer about csrankings; I'll definitely take a look. One more question I had is how you sold the story of going from industry back to school. Was your prior work heavily related to machine learning?

Thanks! No, my prior work was not heavily related to ML. I demonstrated my interests through research and familiarity with the field. To be honest though, I cannot say what aspects of my application were deciding factors. At interview time, I think it really came down to research fit with my POI as I was one of 10 interviewed, for two open slots.

I will say, my application was not the strongest academically, which looks like something you may not have working against you.

Best of luck!

[cabbysaurus]

On 2/25/2018 at 12:54 AM, am_i_too_old_fr_this said:

(I posted this in a few different places, so apologies if you've seen it somewhere else)

 

I'm a software engineer looking to go back to school and get a phd in computer science with a focus on machine learning. I have a BS in a non-CS engineering field and an MS in applied math (where I took a lot of CS classes), both from top 5 institutions in their respective fields. After graduating from my MS ten years ago, I've worked at two startups as a software engineer / manager. The first startup got sold to a large company and the second (current) startup has been gaining traction and I'm now at the VP level managing a pretty sizable team.

 

I do enjoy my job, but I don't love management and I've been intellectually curious about machine learning for a while now, doing a lot of self-teaching on nights / weekends. Originally, it was just to keep up with the current research and see if anything was applicable for work, but as I dug deeper I've found a few areas that I'm very interested in doing further research. I've also saved up enough money that I can afford a grad school salary for a few years and still support my family.

 

Unfortunately, outside of my self-study, I've only worked tangentially with machine learning (mostly using existing libraries and services). I had taken some related classes during my MS, but that was ten years ago. I also don't have any research experience in the field, although I did do some undergrad research during my BS -- unfortunately, no publications and in an unrelated field. Plus it was over 10 years ago.

 

I had a few questions I was hoping to get some guidance on:

 

•  I'm looking to apply for Fall 2019 admission, so I have some time between now and the end of the year to strengthen my profile. Assuming I keep my current job (so I only have nights and weekends free), what are the best things I can do at this point? I had a few projects in mind, but would these help my admission chances?

 

•  I think I can get strong recommendations from my managers at the companies I worked for. They're at the CTO level and well respected in industry, but probably not in academia. I think it'll be a bit of a stretch asking my MS professors since it was so long ago, and they probably won't be able to write anything meaningful at this point. A third option I was thinking about is asking the same professors I got recommendations from when I first applied to my MS -- they knew me well at the time and wrote strong recommendations, so hopefully they still have those LoRs around. Which one of these options would look best to the admissions committee?

 

•  Assuming I can do well on my GRE and my grades are decent (3.9+ both BS and MS), where should I be aiming? I understand the lack of research experience is a huge negative. Should I consider applying to any MS programs with the intention of doing some research and getting stronger LoRs, and then reapplying to phds afterwards?

 

I applied to PhD programs this year in a similar situation. I finished my undergraduate nearly 6 years ago and have been working since, first as a software engineer then as a CS teacher. I also completed a Master's in a related field (Computational Linguistics) part time.

I've been lucky enough to be admitted to a number of great universities (UW, Harvard, NYU to name a few) even despite a low undergraduate GPA. I don't say this to brag - just to let you know that it is certainly possible (as mentioned by @spamhaus as well). Here are the things I think helped my application greatly:

• Research experience - this cannot be stressed enough. I did undergraduate research experience, a quarter worth of graduate research experience, and did a full research internship this past summer. None of this research resulted in publications, but by having done it I showed that I am invested in it and aware of what it entails.
• Letters - I got one letter from my undergrad research advisor and one from a professor who supervised me this summer. The third letter was from my old manager at Microsoft. I would recommend getting one letter from an employer, but the other two should definitely be from academia and the more well-known the better (as long as they're good of course). The letter from an employer should preferably be someone who has a graduate degree and thus can attest to your abilities to do research. They can also mention how great it is to work with you.
• Statement of purpose - I spent SO much time on this. The key was making it unique, showing my passion, and linking all of my past experiences (work, research, and teaching) to my desire for getting a PhD and success in getting it.
• GRE - I did at least average to all of the schools I applied to. Definitely study for it, especially having been out of school for a while. The math is not hard, but the way they ask the questions can be, so going through a lot of practice teaches you the tricks. Also take a lot of practice exams. Sitting for 5 hours straight is hard and I would not have made it had I not done a number of practice tests before hand.

Now to answer your questions:

• I think you really need to try and get research experience. This could be a long-distance project with a professor as @spamhaus mentioned, a summer opportunity as I did, or maybe even something at your job if possible. Without research experience, I don't think you will be a competitive candidate. I also think that you need to be prepared to quit your job in order to gain research experience if necessary. 
• For letters, you really need to get at least two people who can talk about your research. Those people should be professors or researchers with PhDs themselves. I think at least one of them should be recent (within the past year from your application).
• For where to apply, that depends on a lot of things like where you can relocate to, if they have things you're interested in researching, size of the program you want, etc. Some less competitive schools you could consider are Boulder and Northwestern. The most important thing is to find professors who have research that is aligned with yours, then to mention them in your SOP and say why their research interests you and how it aligns with your goals.
• As for getting another Master's, I think that could be definitely be helpful for getting better LORs and some research experience. However, you then have to think about how you would balance that with work and how you would pay for it. I think you may be better off looking for other research opportunities instead.

I hope that helps! Of course, I'm not an admissions expert, just sharing my experience. Please feel free to reach out with additional questions.