What is a realistic TC expectation for a PHD fresh grad in Deep Learning in CS?

Oct 27, 2018 25 Comments

Hello, I am just wondering what is realistic TC expectation for a PHD fresh grad in deep learning in computer science. It seems that the TC can varies a lot in different companies. Even within FAANG, there are many different levels and TC packages.

I estimate that for the low end, it is around "100K + no equity"; and maybe for the high end, it is around "170K + 500K(4 years vested) ". But I am not sure how many people are there in the low end and high end respectively (if 80% of the people are in the low end and only 1% in the high end, I probably shouldn't expect TC be very high).

And I am also wondering that will the big internet companies (like FAANG, uber, airbnb etc) be able to continue to pay stocks to new hires in the future? What is the FAANG, uber and airbnb 30 years ago (are they semiconductor/hardware companies like Intel, IBM, Cisco, HP etc)? And do those semiconductor/hardware companies pay that many stocks to new hires 30 years ago (it seems that right now the total TC in semiconductor companies is quite low and they do not pay stocks)? The reason I am asking is that the stock actually accounts for a very big percentage of TC, and if the internet companies stop paying stocks, a high end TC will only be 170k per year (considering silicon valley is very expensive to live, 170K is very hard to save up anything). So I am just concerned that maybe in one day, the internet company will stop paying stocks, just like the semiconductor companies.

Because I don't know anyone who is in the semiconductor industry for 40 years and don't really know what the TC was like working in semiconductor companies 30 years ago when there was a boom in semiconductor industry. If anyone knows anything about it, can you please shed some light on it? I am very concerned that history will repeat itself and internet companies will eventually become today's semiconductor companies.

I am sorry for the long post. Thank you.

comments

Want to comment? LOG IN or SIGN UP
TOP 25 Comments
  • Nvidia / Eng KXJK44
    Too many thoughts. Just get offers and take the best one.
    Oct 27, 2018 0
  • Amazon / Eng but nvvf
    I guess PhD is over. Time to take precision questions and answer class for you.
    Oct 27, 2018 0
  • Microsoft zizizu
    You think too much. It's not like semi industry workers are living under bridges now. Staying competitive has more to do with the person than the company. Sure, if you keep moving pixels for 30 years you'll probably lose your job to a robot, but you can also keep learning and be the guy who builds the robot. Both are viable paths within many big tech companies.
    Oct 27, 2018 2
    • Microsoft zizizu
      And given you're already in AI, don't worry. 90% of the society would have to lose their jobs before you become non competitive.
      Oct 27, 2018
    • OP
      Yes, I do try to plan ahead as much as possible. Because I am not really sure if deep learning PHD will eventually become a VLSI PHD (semiconductor field). Because I know a lot other VLSI/Analog PHD do not get any stocks even from big companies like intel, Qualcomm, IBM etc (and the base salary is only around 100k in silicon valley). So I am just concerned that the internet companies will becomes the semiconductor companies eventually. Maybe when the internet companies' stock stops growing like most semiconductor comapnies, it will stop paying stocks and TC will drastically drop, is it correct?
      Oct 27, 2018
  • Google / Eng UUKg61
    Depends on your work and experience,if any , before Phd as well. If you are fresh out of college, 150k should be easy and 200k is a great start. After all, a PhD does not make you superman - just that you can go to much depth of a problem. But then , you would need the exact area of work which limits your options.

    In Rome, do what Romans do. If the stock companies are paying stocks, it’s great.If they stop, it would still be good as living costs would go down and it would still be great. Don’t overthink. Maximizing your potential to get the best offer you can , is the best outcome.
    Oct 27, 2018 0
  • Facebook / Eng
    ddx

    Facebook Eng

    PRE
    Amazon, Microsoft
    ddxmore
    just leetcode and interview and see
    Oct 27, 2018 0
  • Realize that if you were in semiconductor industry in the 1970s or 1980s you just have to make 25% more than US average household income which would be about $91k today and Silicon Valley would be your b**** because it used to be much cheaper to live here; $155/sq ft in 1986 in a mountain view Eichler ($170k; 1100 sq ft, 1986; $140k in 1984, 2443 Thaddeus drive 94043) for example. 30 years ago PhD CS started at ~$50k. You could almost buy a house on a starting salary as the valley was probably less than 20% tech workers
    Oct 27, 2018 4
    • OP
      So actually the semiconductor companies were not actually paying big stocks to recruit like the internet companies today? So this is first time in history that the tech companies pay big stocks to hire? Do you think the internet companies will stop paying big stock eventually(maybe when their stocks are not inflating any more)? Thank you.
      Oct 27, 2018
    • Equity comp became a thing in the 80s and 90s. In the 60s and 70s I think it was Founders and maybe 1 or 2 key hires. Big corps (I worked at Xerox OSD) did not give equity. It really took off when they passed a law (I think in mid 80s) that let companies take a tax deduction for the amount that options had appreciated at exercise time. Not only that options were not counted as an expense to companies books until 2005. So (a) employee gets tax-free option compounding, (b) company gives away something worth $x at no cost (originally) and gets a huge tax deduction for employee compensation of $y much bigger than $x many years later. If these laws are ever repealed stock (rsu/option) grants would disappear overnight. https://turbotax.intuit.com/tax-tips/investments-and-taxes/tax-deductions-for-employer-owned-stocks-rsus-stock-options-espps/L0WQuQLD1
      Oct 27, 2018
    • OP
      Thank you very much for the information, Sir. So the semiconductor companies do pay a lot of stocks to hire people in the 80s, just like the Internet companies today. So when did they stop paying stocks? Maybe after the year of 2000?
      Oct 27, 2018
    • I got plenty of incentive stock options in the early 2000s as an employee. These were more lucrative than RSUs because they gave out much more ISOs than RSUs. Especially if the tech stock rocketed up. Say you joined Apple in 2003 with the adjusted stock price at $1.50. If you exercised those ISOs in 2010 at $42, you could buy $42 worth of stock at the $1.50 strike price, multiplied by how many ISOs you had. Of course, you get taxed on the initial paper gain and get hit with AMT, but imagine if you HODL'd that Apple stock for another few years. After the backdating options scandal around 2005 (which Apple was also implicated in), most tech companies only awarded RSUs in lower quantity. That year at the performance review award, I remember thinking what happened to the options that used to be awarded? Why these (relatively) small number of RSUs?
      Jan 17
  • Amazon Northman
    If you are good, target 250k TC. If you are exceptional target 350k TC. If you are OK, target 200k.
    Oct 27, 2018 2
    • IEEE cfgauss
      Agree with this one. 350k is possible but very hard, for really exceptional.
      Oct 27, 2018
    • Amazon mahasan
      Confirmed. I got ~280K and would consider myself very good and from a top school, but didn't have a top conference paper. A teammate fresh out of PhD with NIPS & CVPR papers got 350K.
      Jun 27
  • Based on your publications and citations. I rmb OpenAI pays someone 1M-ish to do AI research (check NYTimes)
    Oct 27, 2018 1
    • New almost5
      That's either kaparthy or goodfellow. They were the 'first generation' students in this new AI wave coming from the big 3 AI labs. And they didn't get 1M out of school, took a few years.
      Oct 27, 2018
  • Microsoft Ink8i
    ~220K is a good target
    Oct 27, 2018 0
  • HP / Eng
    Brodozer

    HP Eng

    PRE
    F5 Networks
    Brodozermore
    HP - About 95-120k TC if you came in as SWEIII in the Portland area.

    Practice being short and concise.
    Jan 16 0
  • PhD School matters a lot. Top 10 can earn 25% more it's like this in all fields. PhD equates to 5y experience.

    Relevance of PhD can also matter. if you were smart enough to compete in the DARPA Grand Challenge you could double your salary.
    Oct 27, 2018 0
  • LinkedIn / Eng
    Gill Bates

    LinkedIn Eng

    BIO
    [Insert epic sax here]
    Gill Batesmore
    You can definitely save money on 170k base salary.

    For TC, it really depends on impact. Do you have a high H index as a PhD student or no? If yes, then low/mid seven figures wouldn't be unrealistic. If you have to ask though, my guess would be somewhere between 120k-260k.
    Oct 27, 2018 0
  • New almost5
    I would target for ~$200k as baseline. You can easily get that with SDE2 lvl at Amazon and MSFT. If you are from an elite group, maybe $350k for AI scientist position at Google or FAIR.
    Oct 27, 2018 0
  • Google csd11
    Stocks are money cause they are now vested in a much shorter time. High-end machine learning Ph.D can get 400k-600k for a fresh graduate. But that represents probably the very top ones.
    Oct 27, 2018 0
  • Amazon Dumps
    200k
    Oct 27, 2018 0
  • Amazon / Eng
    onMyWay

    Amazon Eng

    BIO
    AWS
    onMyWaymore
    ~150
    Oct 27, 2018 0