Meta FAIAR

Amazon
Gkjf16

Go to company page Amazon

Gkjf16
Feb 2 4 Comments

Talking with a recruiter and he is selling me on a Research Engineering team lead role. Does TL imply E6? He describes this as a horizontal applied research group that basically embed themselves into other teams across the company to solve ML problems. I read a couple older things on Blind where people were complaining that you don’t actually do much model work, but rather use pre build models across in some kind of central infrastructure. If that’s the case, it sounds like a nightmare. Anyone that has worked there can give some insight? Any partnerships with FAIR group to bring their research into the real world?

L6 Applied Scientist, TC $310

comments

Want to comment? LOG IN or SIGN UP
TOP 4 Comments
  • Meta
    DidiFB

    Go to company page Meta

    DidiFB
    Congrats first if all, a TL role is nice.

    On embedding with other teams: It will suck. You will need to work with random teams and their random tech stack and get your team’s special sauce deployed in it. 70% meetings, rest engineering, mostly backend and pipelines. No research.

    This could be anything like get teams to be differentiable private, get teams to use uncertainty, … you will have some mostly off the shelf solution here and fight to even get impact with those in the company.

    These things sound good at 10k miles and you see lots of interesting problems but the reality of it will suck quite likely as this typically is not on the innovation side. Meta works a lot like that.
    Feb 2 3
    • Meta
      DidiFB

      Go to company page Meta

      DidiFB
      If you deploy this stuff your career will grow, that’s up to you. Meta is good with that. I am just debating the nature of the job.

      Re off the shelf:
      I don’t know the team obviously, but: this could be a team that says “we build our own library for X” and has some PyTorch library where they’ve implemented the K standard techniques for X. Maybe they will also implement one fancy one, but the team will likely not be innovating on those even if they claim so. The job then becomes to take that library and deploy it against various problems so we can show that X is impactful in products so we can justify investing more in it in the future.
      Feb 2
    • Meta
      DidiFB

      Go to company page Meta

      DidiFB
      Just to be clear: it’s sensible. But I would dig into what “deploying that library X” really translates to in practice. It can end up a never ending fight with incompatible backends and pipelines instead of being a job about the content of what X really is and does.
      Feb 2