Lately I've received many DMs asking about compensations in Europe for ML roles. It seems that comp information outside the US is either badly estimated or hugely skewed by US comps. So I built a simple app last wkd for collecting European tech comps:
Researchers’ idea of retirement:
Open a cafe in Asia, learn a new language, and write papers.
Move to an island, learn fishing, and write papers.
Buy a house in a remote European town, grow your own vegetables, and … write papers.
Got my global talent visa within a week after submitting the application! Yayyy! The github follow count must've impressed them :DD Thanks for the community. If anyone has questions, I can try my best to answer them.
Had a live coding interview in which I couldn’t remember the args passed to conv layers in pytorch. Had to swear to the interviewer that I’d written like 100 cnns. Recently I had to turn copilot off as a preparation for future interviews…
ChatGPT and copilot have completely changed the way I code. I now spend much of my "coding" time giving detailed instructions and copying or accepting generated code. I type way less and look up documentation way less. There's no going back.
@desired_state
Current situation is companies asking applicants if they need a visa as part of the application so they can filter them out like this, without specifying if they can sponsor. Why wouldn’t you say it in the job description so ppl wouldn’t waste their time to apply? Seems sane too.
🎉My quickest job rejection record: I submitted my application 10pm the night before, and got rejected at 9.07am the next morning. It's like someone went to work, made a cup of tea, turned on their computer, and BANG💥!
Holiday’s over! I’m starting today
@sourceful
as an ML tech lead to improve the package design platform 🎉🥳🙌 I’ve been wanting to work on generative models for design since the first ever GAN project. So excited that this journey finally started 😃
@MarkMoyou
Oh no, would I be blacklisted by NVIDIA now?
If skill sets for doing LLMs are required for this role, there are more relevant bullet points than minimum 8 papers.
Having missed two NeurIPS due to visa issues, and two more due to the pandemic, I'll be attending my first in-person ML conference ever after an entire 4.5 years of PhD life. How exciting 🤩🤩🤩
While some worry about not having a publication, I keep getting rejected for being "too researchy". You will have to be a full stack research scientist to stand a chance in this market.
Recruiter: when was the last time you helped someone solve a technical problem?
Me: last night. I helped ChatGPT fix the bugs in their code.
The recruiter didn’t laugh.
@Charbel_Sakr
They are looking for those at a large lab that puts everyone’s name on papers. I know great academics who only keep 1-2 students over a 4-year span. The quality of papers produced from them is high. What’s the point to “collaborate” just to get 8 papers in 4 years? It’s nuts.
@zornsllama
@kgourg
I don't know anyone who published more than 8 papers to all of which they made substantial contributions. This kind of requirements just favour candidates in a big lab.
Glad to see discussions on leetcode interviews following my failure last Friday... Just want to give a bit of context, based on my personal experience.
I mainly apply for AI/ML scientist roles, mixed with a small portion of MLE but never SWE.
I passed my PhD viva today! Big thank you to my examiners Prof. Milan Vojnovic and Dr.
@shakir_za
for a great viva experience, for the opportunity to discuss my research, and for the insightful feedback. Never thought I would enjoy it so much 😆
London housing market is in such a crisis that I used my Dr. title for the first time in case it might add any bonus points to my offer. For those who are considering doing a PhD for the title, this is today's motivation boost.
Rejected by
@icmlconf
. True that the review quality isn't great. Also true that the paper doesn't deserve to be published. I accept that and am moving on to the next thing. Just to put it out there that this is also an option.
Let me get this straight. You can use a T4 on Colab for free for 12 hours without subscription. But once you subscribe to Colab Pro with compute units or buy them via pay as you go, T4 costs about 2 units per hour to run?
I’m on the job market again, planning to come back to generative models. If you know any openings in this space, a tip would be highly appreciated. My personal website is in the bio, and google scholar page is here:
Sam Altman was hailed as "the father of ChatGPT" by the Chinese media. Given how little fathers usually contribute to parenting in China, I don't disagree.
Pivoting my influencer career to be more ML-serious. How about I review and recommend papers like
@_akhaliq
but the worst ones instead? Give me some ideas.
If you are applying for the UK global talent visa as an AI/ML expert under the digital technology sector, you cannot show off your proficiency of using any AI/ML technology to make this application.
I think I think! That they are trying to make a joke.
@isabelacarv
@ghostapple_ari
@cindywangbrandt
As someone born and raised in China, I totally agree with this. Specifically calling a friend to go to the supermarket, dry cleaning or the tailors to alter your trousers is pretty weird to me. It probably depends more on the person rather than nationality.
10 years in the UK, still struggling with military alphabet. Whenever I need to spell my name on the phone, all the words start wrestling and flying around. I was like T for transformer, I for intelligence, A for artificial .. N for ummmmm nan
IMHO, it'd save both of us some time if you wrote a little sentence about this in the 2-page long job description. But thanks for contributing to my rejection collection.
Random person: “I don’t like the method proposed in the paper. But I like the paper.” 🤯
Can we pleeeease all agree that if you don’t like the patty in a burger, it means you don’t like the burger?!
3rd week into my internship. I'm convinced that anyone who can sit in a 2-hour long meeting and still ask questions about the technical details is from a special talent programme I do not belong to.
Oh yea how disappointing when I can only earn more than 95% of the population instead of 99% at 22 years old. The economy is bad. People are struggling to have their ends meet. And here's us, complaining about not being able to get FAANG jobs so easily anymore.
We’ve got over 100 entries now for tech salaries in Europe, including FAANG firms as well as startups. Thanks for those who contributed! It’s not a lot of data but should provide a rough idea for what you can expect.
Lately I've received many DMs asking about compensations in Europe for ML roles. It seems that comp information outside the US is either badly estimated or hugely skewed by US comps. So I built a simple app last wkd for collecting European tech comps:
Random DM: wow can’t believe even you are struggling to find a job. You can imagine what it is like for a man in this market. Blah blah blah, but you are a woman, blah blah blah blah
Which year is this? Pisssss ooooooff 😠
I started my job search by leetcoding. But in my recent interviews, 80% didn't ask LC Qs. Could be because most of them are with smaller labs or startups. So I stopped spending time on it. The difficulty of the Q asked in my interview was reasonable. I am just bad at it.
Every time when I post a ridiculously low salary, US employees are like, ahh it’s the UK, poor British workers. Lololol no it’s not. London tech salaries are often about 50% of what you get in the states, not 10%, okaaaaay?
@linylinx
Allow me to apologize. This bullet item was posted in error. NVIDIA looks for the quality of publications. We are not concerned with quantity.
@nirsd
They asked me to implement a CLIP style model in a live coding pad that doesn’t have an auto-complete functionality. To be fair, interviewer wasn’t too bothered that I couldn’t remember the args. But I’m worried the next one wouldn’t be so chilled.
It was much more common for the first round to be leetcode interviews in the past, especially at large tech companies and investment banks. Normally they are easy-medium level questions, although banks really like to ask DP problems.
Ugh. Who told me nobody would ever read your PhD thesis except for your supervisor and examiners? If I knew it'd be cited in some sexy foreign language, I would've put more effort
Wow somehow I've been completely erased from the history of Kaggle.
For the record, the US company that was funded, Kaggle Inc, was owned equally by Anthony Goldbloom and I with a minority holding by Nick Gruen. Ben Hamner didn't join until later.
Was horrified by some tweets on my timeline. Thought this site was becoming LinkedIn. Then I realised it was under the “For You” section. Phew. What a shit feature.
@JWach26
@thechosenberg
You speak as if phds can’t be cs grads lol if you interview a PhD who have worked on theoretical work only for a technical role, that’s on you.