Ask HN: Good engineer, bad grades in undergrad. How do I get back into academia? | Hacker News Reader

Ask HN: Good engineer, bad grades in undergrad. How do I get back into academia?

18 pointslosing_hope18 comments

Dear HN,

I just turned 24 and am having a quarter-life crisis.

I graduated over 3 years ago with very average grades from a decent university by doing badly even in the subjects that I knew I was great at because:

1. I wasn't good at writing timed exams (I just cannot think under massive stress) 2. On-and-off phases of depression

Fortunately I am a very good engineer and I did a bunch of open source work while at univ and after graduating have been almost exclusively writing FOSS for the past 3 years. And not simple CRUD apps, but difficult engineering problems.

But here's the thing:

I enjoy writing software, but I don't see an exciting future for myself doing the kind of work I do. I have a great aptitude for Math/CS and I want to do something that is more exploratory research (currently obsessed with ML and specifically Neural Networks). I want go back to academia because 3 years of working in the industry doesn't give me the kind of satisfaction that university research would. I want to broaden my horizons; I'm also getting older and with each passing year it will get more difficult for me to resume academics. I keep flirting with interesting courses on Coursera etc, but deep down I know it's not working because it genuinely doesn't compare to what I would get in a proper degree program. Ultimately I want to get a PhD and I'm willing to work hard for that.

But it looks like all the good places want someone with at least a 7.5 CGPA (on a 10-point scale), and recommendations from professors. Mine is only 6.5 and I expect to get recommendations from (high profile) industry figures.

I'm Indian, but based in Germany, and looking for univs in Western Europe. I don't believe I could ever afford American ones.

So I'm asking for help, HN:

1. Do you know if admissions committees can make exceptions for people with demonstrated open source work? 2. Any professors/academics here who may be able to help me out?

>I don't believe I could ever afford American ones.

Most PhD programs at American universities cost nothing for most students, i.e., if the university accepts you into a doctoral program it almost always comes with full tuition remission (i.e., you pay nothing) and a modest stipend (i.e., they pay you, somewhere between $10,000 and $25,000 per year). The stipend can be in the form of a teaching or research assistantship, in which case you will work 10-20 hours per week helping a professor, or in the form of a fellowship which requires no work (but fellowships are highly competitive and you likely would not receive one).

Does that change if you're an international student? I honestly don't know, but it would surprise me.

The same funding applies to international students, but the admissions criteria is a little steeper for foreigners. You also need very good GRE/ english proficiency test scores to get in as an international student.

A PhD especially in ML is not about Software Engineering or learning how to write code. Learn how to do research, write a proposal, survey the breadth of the field. Find interesting problems. Attempt to get to the stage where you either publish or are able to have an interesting conversation with your professor.

There are multiple ways of doing this.

1. Do this solo.

2. Work in a research lab as an intern.

2. The other possibility is to do a Masters and work with a Professor.

Getting into a PhD has very little to do with your GRE, undergrad grades or where you went to school from. These are merely supposed to be proxies for your ability to focus on doing research. I would argue that you should spend sometime doing research to understand whether you actually like doing research.

I will tell you this... it is a real issue. I have a similar issue in life as I get bored with a teacher standing in front of me talking theory, and I too deal with the depression...

My advice to you would be looking into certs... I have been doing alot of self education and I believe certs will give you not only the proof that you know a topic but the feeling of completeness.

I am 23 I will be 24 this month on the 28th, I have been actively programming since I was 12 and I have been doing it for a living since before I graduated high school in 2010. I recently began taking courses at my local community college to bring up my GPA which was a 1.9 when I left due to some of the same reasons you didn't do so well. I lived to prove that you can be successful without a degree, my family told me I would never amount to anything without a college degree, and I am making around 125K in Detroit, MI where the median salary is 35K.

With all of that being said, just take the small steps in the right direction, I know I have and its a good feeling.

I hope this helps.

>certs will give you not only the proof that you know a topic but the feeling of completeness.

Do people care about certifications outside of IT/Sysadmin jobs? I don't think CS research groups and/or professors even look at them.

I am not sure what the situation is like in Europe, but (coming from the US) I think that your conception of what the life of an academic is like is probably tinted by rose-colored glasses.

At a good university, you will be expected to serve on numerous soul-sucking committees, your students will be sucking up to you so that you will hook them up with good recommendations and/or good jobs (and you better deliver), and your dean will be riding your ass to get grants, publish, and get rated highly on poorly designed course surveys.

At a less-than-good university, you will still have the course survey problems, but you will largely have a student body that is best described as apathetic and untalented (there will be a few good students, but they will be the exception rather than the rule). You will also have a larger teaching load. Good luck getting a good research grant -- I hope you have a great professional network. Expect your research to be reviewed a bit more strictly, because no respectable journal wants to publish an also-ran.

Now that the reality check is complete, let me make some pro-active suggestions:

- If you are any good at all at ML, there are many organizations with interesting problems who will snap you up in two seconds flat. You just need to be good at ML and have a way to demonstrate that.

- As far as learning goes, find the people you want to work with and figure out a way to work with them. There are a handful of researchers who have a surplus of talented people who want to work with them, but most researchers are starved to find people who are both talented and passionate about their subject. If you can be that person, they will find a way to make it work. It might be on a grant, it might be as a grad student, it might be something else -- just roll with it.

- If you can't make Coursera work for you, then I truly wonder about your motivation. What do you think you will get from a "proper degree program"? Most of your professors (at good universities, at least) will not want to hold your hand -- quite the contrary. From my perspective, the best grad students go to professors for suggested readings and to answer questions when they come to a (relatively rare) road block in their autonomous learning. Outside of that, peer networks are peer networks.

I think this is enough for now -- feel free to ask follow-up questions.

As a side note, I get the sense that some of your issues/challenges are being fed from the psychological side (e.g., what does it take for you to be "satisfied"). I suggest you talk to a professional if possible. The road to self-actualization is often riddled with potholes.

This is math/CS. You don't need a particle accelerator to generate results, just a pencil, paper, and a computer. Do research, get published. That's the currency that matters in academia, and that is what will make a professor ask you to be in their program.

If your FOSS work has publishable results, publish it.

I think you are massively underestimating the difficulties to get published in a decent journal.

At this point peerj will be fine. Peer reviewed is what matters at this point. Also find a Prof whose work you like and are interested in. See if you can use some of their published work to do something neat.

Is it actually easy to publish work outside a lab?

Did you get a master degree already? If not, it should be fairly easy to get into a German university to get that and go on to do a PhD. Have you already tried to apply? Which university? If you can't get into a top university, try _any_ university and move to a different one for your PhD. The best German universities will likely accept you for your PhD, if you do really well while getting your master degree, no matter where you got it.

Being good at Math and writing open source software plus having an intense interest in ML, you might try contacting a few professors working in the field you are interested in and volunteer to write up some software in exchange for some help.

You can contact them online, but try contacting 10 or so, not just one and getting disappointed if you get a bad response.

Take a look at some problems on Kaggle. Enter a competition, see how you do, publish your results or at least use them in your application package. You can also find startups in this space (some are in Paris). You might also find a job at Google/Facebook. Then you might be able to get into grad school as well. Try all the options.

I'm in a similar boat. My plan is to learn ML through application. I'm going to start doing exercises on Kaggle.

I'm also going to start focusing on the computing/architecture side once I get the theory side down.

I wonder how many other folks out there are in similar boats? We should band together lol.

Hmm, at the risk of self promoting, I would be happy to have a chat with you. I like helping students out and do it professionally as well. You can read more and if you would like to chat, please check out the contact page. My website is www.scholarstrategy.com

How good is your German? German universities are practically free, so maybe you could join an undergrad program to get inside the network?

I know a friend who did this for phd.

Just apply. See what happens.

you should come to india first, we have many jobs for expert coders here, u could land up getting u a self quite many opportunities if u have good coding skills...

If I would have been a company owner and if u have had that skills i would have taken u even if u had that level of skills \\

Reply on news.ycombinator.com