5 ms·

Someone on another thread here about hiring in the ML/AI space bemoaned the fact that they mostly get applications from people who have dabbled with ML/AI on th

Someone on another thread here about hiring in the ML/AI space bemoaned the fact that they mostly get applications from people who have dabbled with ML/AI on the side and very few applicants who have actual ML/AI work experience. But it seems like the ML/AI space is still nascent and growing rapidly - you're not going to find a lot of available people with ML/AI work experience. Given that ML/AI is where a lot of current hiring is maybe that's the reason for the hiring disconnect? If so hiring managers are going to have to start looking at people who have an interest and have done some self-learning in the field.

jwells89 3y ago
It's another variation on the classic, "6 years experience with X required" where X is something that has only been publicly available for a year or less. Everybody wants candidates who can hit the ground running even when that's not practically possible.
- atrettel 3y ago
  This is just one example of the overall problem. In my view, hiring managers often are just going off a list of requirements and do not really know too much about the technology. They tend to reject anyone who does not use the exact terms they are looking for, even if they have a sufficient amount of experience. I'm thinking about people who worked in optimization for a long time before applying to machine learning jobs. Both fields have basically equivalent work, just with different techniques, and optimization is a lot older and more mature of a discipline. If hiring managers do not know that, they will skip over qualified candidates.
- throwaway5959 3y ago
  That sounds more like H1-B shenanigans. Easier to justify not being able to find someone local if they make the job req impossible.
thih9 3y ago
> If so hiring managers are going to have to start looking at people who have an interest and have done some self-learning in the field. Or candidates who have just started working with ML/AI are going to have to accept lower pay as their cost of changing specialisation. Or, likely, both.
- PeterisP 3y ago
  I think there really aren't very many job positions that try to hire candidates who have just started working with ML/AI for a lower pay, everyone wants 'all or nothing', exclusively trying to hire people who already did that specialization change somewhere else; many are willing to pay for experience, but they aren't willing to accept lack of it for less pay.
  - throwaway5959 3y ago
    This is true in general. The average software developer is doing the jobs of three or four people when I started fifteen years ago. Ops, dev, DBA, storage (s3, ebs, etc).
Cthulhu_ 3y ago
The problem there is that a lot of people follow hypes because they think it's cool and/or earns them a lot of money. I've also seen it in job adverts though, at some point pretty much every software development job would have "crypto" and "IoT" sprinkled in the description. It's tiring. And it's insulting. And also, to the developers that actually look for that kind of thing without already being in that space, have some pride. It's a Java job, you aren't going to turn your home climate control hobby into a job through that one.
quickthrower2 3y ago
“We need 10 years GPT experience!”
jwestbury 3y ago
I used to work in the cleared space -- i.e., people with top-secret security clearance. My first job in this space was at AWS, where we probably had about 1/4 of the qualified, cleared talent pool in our space. It was almost possible to hire people with sufficient technical skill who also had clearance -- but it was equally hard to get people cleared, because we and Microsoft were pushing so many people through the pipeline that clearance decisions were often taking 2+ years. We would have people get hired, get a promotion, and leave all before getting their clearance decisions. My team started taking a different approach: We started hiring people at the lowest level we could ("associate", below even entry-level engineer), with minimal tech skills, but with active clearance and some adjacent skills. STEM background, or tech writing, or just someone who was a hobbyist. I started interviewing people on troubleshooting -- someone told me a fantastic story about troubleshooting a comms issue while deployed in the middle east. The fix involved copious amounts of superglue on a ground-based satellite. Switch out the words, and you could easily have a story of a grizzled old sysadmin patching in production. We actually ended up being pretty successful. Sure, a substantial portion of our hires were still misses -- but that was already true. Meanwhile, we hired some really talented people from this route, who I ended up trying to poach years after leaving AWS, and leaving the cleared space entirely, because they were just genuinely great engineers who needed to be given that chance. tl;dr, I agree, and it's nearly always a good idea to look at expanding the talent base in some way or another if you're struggling. In many cases, it means hiring people who have potential, as long as you have sufficient numbers of senior staff to help get them up to speed.

jwells89 3y ago

It's another variation on the classic, "6 years experience with X required" where X is something that has only been publicly available for a year or less. Everybody wants candidates who can hit the ground running even when that's not practically possible.

atrettel 3y ago
This is just one example of the overall problem. In my view, hiring managers often are just going off a list of requirements and do not really know too much about the technology. They tend to reject anyone who does not use the exact terms they are looking for, even if they have a sufficient amount of experience. I'm thinking about people who worked in optimization for a long time before applying to machine learning jobs. Both fields have basically equivalent work, just with different techniques, and optimization is a lot older and more mature of a discipline. If hiring managers do not know that, they will skip over qualified candidates.
throwaway5959 3y ago
That sounds more like H1-B shenanigans. Easier to justify not being able to find someone local if they make the job req impossible.

thih9 3y ago

> If so hiring managers are going to have to start looking at people who have an interest and have done some self-learning in the field. Or candidates who have just started working with ML/AI are going to have to accept lower pay as their cost of changing specialisation. Or, likely, both.

PeterisP 3y ago
I think there really aren't very many job positions that try to hire candidates who have just started working with ML/AI for a lower pay, everyone wants 'all or nothing', exclusively trying to hire people who already did that specialization change somewhere else; many are willing to pay for experience, but they aren't willing to accept lack of it for less pay.
- throwaway5959 3y ago
  This is true in general. The average software developer is doing the jobs of three or four people when I started fifteen years ago. Ops, dev, DBA, storage (s3, ebs, etc).

Cthulhu_ 3y ago

The problem there is that a lot of people follow hypes because they think it's cool and/or earns them a lot of money. I've also seen it in job adverts though, at some point pretty much every software development job would have "crypto" and "IoT" sprinkled in the description. It's tiring. And it's insulting. And also, to the developers that actually look for that kind of thing without already being in that space, have some pride. It's a Java job, you aren't going to turn your home climate control hobby into a job through that one.

quickthrower2 3y ago

“We need 10 years GPT experience!”

jwestbury 3y ago

I used to work in the cleared space -- i.e., people with top-secret security clearance. My first job in this space was at AWS, where we probably had about 1/4 of the qualified, cleared talent pool in our space. It was almost possible to hire people with sufficient technical skill who also had clearance -- but it was equally hard to get people cleared, because we and Microsoft were pushing so many people through the pipeline that clearance decisions were often taking 2+ years. We would have people get hired, get a promotion, and leave all before getting their clearance decisions. My team started taking a different approach: We started hiring people at the lowest level we could ("associate", below even entry-level engineer), with minimal tech skills, but with active clearance and some adjacent skills. STEM background, or tech writing, or just someone who was a hobbyist. I started interviewing people on troubleshooting -- someone told me a fantastic story about troubleshooting a comms issue while deployed in the middle east. The fix involved copious amounts of superglue on a ground-based satellite. Switch out the words, and you could easily have a story of a grizzled old sysadmin patching in production. We actually ended up being pretty successful. Sure, a substantial portion of our hires were still misses -- but that was already true. Meanwhile, we hired some really talented people from this route, who I ended up trying to poach years after leaving AWS, and leaving the cleared space entirely, because they were just genuinely great engineers who needed to be given that chance. tl;dr, I agree, and it's nearly always a good idea to look at expanding the talent base in some way or another if you're struggling. In many cases, it means hiring people who have potential, as long as you have sufficient numbers of senior staff to help get them up to speed.