Taro Logo
0

Rate for training data as an engineer?

Profile picture
Senior Software Engineer at Taro Community2 months ago

I got some random DM on LinkedIn asking to recruit me to train their data for their AI company at $75 an hour. Is this really a good rate if I have to write prompts and think? They said they valued PhDs most, but others were strong candidates with MS and BS were ok too. I just wonder if it's worth the time contracting vs working on code. It just feels like data munging/mechanical turk...even if it says it requires this domain expertise to stump a foundation model.

11
1

Discussion

(1 comment)
  • 1
    Profile picture
    Tech Lead/Manager at Meta, Pinterest, Kosei
    a month ago

    interesting! this is a very new space so I don't think there are norms around pay yet. DM on Slack and I can share more opportunities or provide some commentary. In a vaccum, $75/hour is reasonable but I've definitely seen the hourly wage for SWE work be much higher, like $150/hour.

    Pros of AI gig work:

    • Flexible, remote-friendly
    • Part-time, usually 10-20 hours/week
    • Easy to stop anytime, and you won't 'let the team down'
    • Decent hourly income which can be a good supplement to your day job

    Cons of AI gig work:

    • The work is often not that interesting, just doing some analysis/preference selection. This will change as the tasks become harder/more involved as LLMs get better.
    • Unstable work, hard to predict how long a project will last
    • The pay rate is variable
    • There's no concept of career progression, you likely won't get much mentorship
    • There are no benefits, e.g. an office or health insurance

    Is it worth your time? I would say yes if:

    • You have 10+ hours of free time/week and the extra income is interesting for you.
    • You want to understand what goes into training AI models