India
Yesterday
1233
Ideal indian parents
Software Engineering Career
4h
1566
L4 Google -> 45 interviews, 5 offers, AMA
Work Visa
Yesterday
1125
Hypocrisy of Indians
India
Yesterday
697
A list of ethnic slurs on Indians that should be banned on Blind
Tech Industry
Yesterday
1097
PM is irrelevant role and will die in next 2-3 years.
I'm a junior in college and really enjoy studying / working with compilers. Is it a feasible domain to break into without a PhD? What companies should I be targeting specifically? I'll be interning with Microsoft this coming summer but on Azure, so nothing compilers related. I see lots of placing looking for ML compiler devs (more senior roles), but I don't know any ML. Is it worth learning if I'm interested in the compilers space? Looking for any advice, thanks. #compilers #microsoft #waymo #intel #nvidia
Same question but for an experienced data engineer (this no experience in compilers): how do I get into a similar role? I’m working on some projects and would like to go as deep as I can, but should I take a PHD of a mid-tier school (in Canada so maybe McGill or UT) if given the chance?
The problem is that 95% of a compiler is very straightforward but the last 5% (optimization, the interesting part) is very advanced.
Is the suggestion that I should get a PhD in order to work on the more advanced parts of compilers?
Either that or a longer journey in the industry. It is a very good skill though.
I got an ML compilers job recently with no prior experience and no MS/PhD and it is definitely an up hill battle. Most roles, even the new grad ones, require prior experience in compilers through a prior job or through an MS or PhD. Another thing you can do in the absence of this is make contributions to LLVM, IREE/XLA, etc, maybe make your own toy compiler to demonstrate some interest, etc. Getting into ML compilers as a new grad is slightly easier than experienced hires since new grads get to team match at a few places, which allows you to join teams and orgs without prior experience. A lot of roles basically just want you to be smart and know about computer architecture and PL, which is how I ended up getting my role (and a very strong referral from a friend working there, every other place I applied to rejected me). As other people said though you won't really be doing the "interesting" stuff until you have a few YOE under your belt in the compiler space and/or you did prior research in an area as a PhD.
Dude, that's pretty awesome. You mentioned no prev experience. Did you just work on PL/compiler related projects to show demonstrated interest? In hindsight, would you say getting a masters in something like systems (with a focus in PL/compilers) would have been worth it? I'm worried about just heading straight into a random domain (like web dev) and getting lazy to look for something for tough like compilers. I really do want to do tough, interesting work, but just worried if I'm not immersed in it directly after college, I'll get pigeon-holed into something that's not my passion and just grow complacent.
My prior "experience" was low level focused internships and TA experience in school. My referral definitely helped me out here by giving me a chance to highlight my strengths, since I talked a lot about my interests and the things I do outside of work (reading comparch papers, publishing open source libraries, etc). They basically made sure my name got in front of the hiring manager, and from there my whole profile made me look decent enough to the point where they'd take a chance on me, even though my work might have not been directly compiler related. I do think it will become harder to get in this way the more experienced you get though, but im relatively low YOE so I could spin it in this way. To be honest my current job is not relevant at all to compilers/low level sw/my interests, and I definitely put in a lot of effort outside my job just studying and getting good so that when I actually got an interview, I'd be sure to crush it. I guess this personality might vary person to person though. I basically viewed my studying as the one ticket I had out of my current job and into one I'd actually like so I put in a lot of effort. I think doing an MS would increase your chances drastically, as would doing PL/compiler research in undergrad, as would making OSS contributions to various compiler frameworks and tooling.
Yes. No one thinks about compilers. Thats where the real bottlenecks are. Lot of work is needed to extract performance out of hardware. Especially with so many custom architectures coming up for ML hardware the need for good compilers folks will increase.
So I think that falls in the realm of hardware accelerators, yeah? That was definitely on my "things to learn". I was thinking of going through this: https://inst.eecs.berkeley.edu/~ee290-2/sp21/ Thoughts, if any?