Amazon AI

New / Data 3blxqv
Feb 10 18 Comments

How reputable is Artificial Intelligence at Amazon within the industry?
Does anybody know about how Amazon AI Labs or AWS AI is?
Any insights on projects, teams, wlb, tc anything would be appreciated.

comments

Want to comment? LOG IN or SIGN UP
TOP 18 Comments
  • Amazon lolwhat
    They're years behind open source counterparts, and in AI space, that's equivalent to eons
    Feb 10 5
    • New / Data 3blxqv
      OP
      Appreciate it, but the AI works applicable within amazon sounds interesting enough? Is that just a mirage..
      Feb 10
    • Amazon lolwhat
      Yes. It is actually one of the better places to work within Amazon.
      Feb 10
    • Amazon JSnowflake
      Not sure what specifically you meant about years behind open source counterparts? What product specifically are you talking about?
      Feb 10
    • Google nooglr101
      Compare ML frameworks - TensorFlow and PyTorch. Amazon just adopted MXnet work done by Tianqi et al.
      Feb 10
    • Amazon JSnowflake
      MXnet is better at production usage and scales like a beast in distributed training. PyTorch wins hands down for pythonic expressiveness and ease of optimizing for crazy loss functions, dynamic computation graph manipulations etc. Tensorflow wins on install base and library code but has had terribly unstable APIs and horrible versioning discipline. Gluon addresses some/many of MXnets deficiencies relative to Pytorch though. Anyway, point is I don't think years behind is fair for DL frameworks. Other parts of AWS AI, maybe.
      Feb 10
  • Amazon queop
    Sucks. Team sucks. Too much politics
    Feb 10 1
    • New / Data 3blxqv
      OP
      Thanks could u please tell more? Is this particular to AI or general across?
      What kind of politics play within AI?
      Feb 10
  • Amazon / Eng Nestle
    3rd rate technology
    Feb 10 0
  • Amazon TC3000
    Tons of AI teams and they run the gambit. MT is currently world class.
    Feb 10 4
    • New / Data 3blxqv
      OP
      MT = Mechanical Turk?
      Sounds like a lot of experimental works, is that bad thing to do in Amazon?
      Feb 10
    • Amazon TC3000
      Machine translation. You have to consistently deliver results at Amazon. They won't wait five years to see progress
      Feb 10
    • New / Data 3blxqv
      OP
      I'm assuming AI work is more experimental at Amazon now than other established products like Prime, Alexa etc.. does that implicate worse wlb, politics etc in your opinion? Which other team at Amazon would you pick if you had a choice?
      Feb 10
    • Amazon TC3000
      Really depends what you are good at, ton of good groups around from personalization, app development with ml, infrastructure, music, a lot in Alexa, and all those secret projects
      Feb 10
  • Amazon jkekxna
    Well, I researched little bit while looking for teams internally.

    - amazon recommendation team sucks big time. You must have noticed yourself that it keeps on recommending the stuff you already have bought or visited.

    - Alexa org is bloated. Multiple teams are doing same thing. There is no vision, and lot of tech debt.

    - AWS AI, it's a managed service, so not much is done at ML. It's about creating a service. So nothing much different from other AWS services.

    - mxnet. Amazon recently hired lot of PhDs and big shots like Alex smola to create something like tensorflow. But it's a very shitty framework. Lot of teams are being forced to use mxnet instead of tensorflow.
    Feb 10 0
  • Amazon Codarnage
    There’s no such thing as “AWS AI”, and “Amazon AI” isn’t a service. Just FYI.
    Feb 17 0
  • Google nooglr101
    Amazon never desires hiring the best A.I. talent. They have adopted MXnet - developed by grad students from some top schools. Particularly they adopted other works of Tianqi Chen (from UW) - most notably TVM.
    Feb 10 0
  • IBM jTkM01
    Following
    Feb 10 0

Salary
Comparison

    Real time salary information from verified employees