Tech companies have a standard coding interview process for hiring these days including several screens and then onsite coding interviews for ALL jobs where any kind of coding is involved. I believe I have some rare combination of skills in deep learning research (no, not via coursera or online blogs, actual research and published experience), cloud computing, and dsp. In day to day work of an ML engineer, we rarely work on optimizing code runtimes. But while applying to any good tech company, they just ignore all that and 90% of the interview is focused on the same code optimization questions that’s asked from anyone who just came out of school. Needless to say, a good coder and good ML practitioner are totally exclusive.
So that they can verify the candidate can actually code. Too many bullshit artists.
This.
Plenty of BS artists that can pass code interviews...
Premise is not true
In the two ones I did recently there was at least a pure coding exercise. Gotta pass that bar 😀 And in some, strong elements of coding for the system design questions.
Sorry but what’s up with the 6/7, these are levels? How many years of experience would you think transition you to that senior level? 5 years should be enough I suppose.
So true. My last Google interview, I showed them how I can burp "Archbishop of Canterbury". Didn't get the job.
Is that a sex act?
Cuz the interviewers are unsure of their own ability to judge other people. So they need a binary decision making process. Oh, you couldn't write an optimal answer on whiteboard in 15 mins? Binary decision no. Doesn't matter that you proved theory of relativity in another round.
Very true. Personally, I take it a step further. They are still human, and the decision they make is still based on their perception of you. With good people skills you can flip the interview and spend time interviewing them instead, build rapport. Once they feel like you’re a part of the team, you can get away with murder.
Even Google Directors, VPs, and Execs?
Well for ML many are/were research scientists, CTOs, ...
Made similar post. Bummer right ML/DL interviews https://us.teamblind.com/s/SgrZNyms
When someone says I have a rare skill, I laugh. Unless your skill is mining diamonds in Jupiter, it ain’t rare. Also if you don’t know how to go to Jupiter (=coding) then your mining skill is useless. All your ML rare skill is just BS if you cannot translate your ideas to reality. Unless you think your job is to say big things and someone else will do the actual implementation. Talk is cheap. Also a good cider and good ML engineer is exclusive? On which world? Jupiter I guess.
I think you read too much into my attempt at separating ML and coding. DL people do actual implementation and all, it’s just that their problems have not much of an intersection with the ability to write all sort algorithms in 15 minutes. Even people in biochemistry work on ML code. Electrical engineering folks are also at a big disadvantage here when they have to compete with CS majors. There’s a lot of ML/DL used in EE, all I’m saying is that while interviewing, the main focus should be adaptive to the job you apply for mainly, and candidate’s background. Maybe they are good at a lot of other skills like data preprocessing. It’s like asking everybody Jupiter questions even if they come from Mars.
If you really have rare skills and published research or a meaningful thesis, they’ll get a PhD/research scientist to drill just into that. But that’s *one* interview.
Its not that hard, a couple of months of practice and you can code fluently in the interviews. If you want the big names you have to put in some time.
Can't agree more. Passing tech bar is not rocket science. Not sure why people fret about it so much.
Yeah I have taught a group of high school students going from zero knowledge in programming to passing interview bar in 4 months. They are talented and top students for sure but they were just having fun. People here are making their living of coding and they don't even want to put some effort into it.
Rare skills are under domain knowledge area.
How about coding being a skill under the domain of ML? Maybe we need to wait till ML can code for us as well :p
What's does that mean? Rare skills under domain knowledge area?