4 ms·

As a YC company that is currently hiring, yes! And all of the YC companies I know are also struggling to find engineers. But the job listings (HN, WorkAtAStartu

As a YC company that is currently hiring, yes! And all of the YC companies I know are also struggling to find engineers. But the job listings (HN, WorkAtAStartup) practically never bring in good candidates. A few big problems: 1. AI Spam. I categorized the inbound we got the other day from a job post. Out of 172 daily applicants, we got 22 that looked reasonably like a person, and 150 that were primarily AI generated messages. Which are pretty easy to spot because they're 500 words of tech jargon and rehashing the job description. 2. Purely automated applications. There are a lot of "Apply to 1000 jobs with AI" startups out there that just spam job boards [1][2][3]. 3. Qualifications. There is a shocking number of engineers applying to work at an AI company who have never made a single API request to OpenAI. After three years of hearing about AI every day if you've never tested a single inference API then why are you applying to an AI startup. The signal to noise ratio is so bad that it's better to just do outbound. At this point the job listing is mostly there so we can share it with candidates that we reach out to. [1]https://lazyapply.com/ https://lazyapply.com/ [2]https://aiapply.co/ https://aiapply.co/ [3]https://www.reddit.com/r/GetEmployed/comments/1eo8uyp/i_used_ai_to_automatically_apply_for_1000_jobs/ https://www.reddit.com/r/GetEmployed/comments/1eo8uyp/i_used...

nedwin 2y ago
Adding to this: we mostly focus on outbound, mostly using WorkAtAStartup or LinkedIn as a source.
rors 2y ago
I have a PhD from one of the best universities in the world in Machine Translation, I was training feed forward networks in 2014, built models in TensorFlow in 2016, founded a generative AI startup in 2019 and signed deals with huge consumer goods companies, debugged distributed training jobs on a cluster with thousands of A100s, and wrote PyTorch training pipelines for a financial ML trading signal that made money. Yet, I am unqualified to join an AI startup because I’ve never made an API request to OpenAI
- yapyap 2y ago
  Make ur first API request to OpenAI today and who knows, rooting for ya!
- mikhael28 2y ago
  This is a bad faith comment - if you had all that in your resume, they wouldnt disqualify you for not ‘sending an api request’.
- beambot 2y ago
  The attitude underlying this comment likely makes you a bad culture fit...
- sshine 2y ago
  Having a PhD vs. making that API call might ironically suggest you don’t emanate hands-on product thinking. I know PhDs in data science and language technology who are not startup material because they’re too uncomfortable outside of MatLab/Jupyter/R.
doctorpangloss 2y ago
The unspoken problem is that most people want way too much fucking money.
- themanmaran 2y ago
  So far we haven't had a candidate pass because of salary yet. Typically candidates looking for that stable $400k salary never even talk with startups, because they know that's only a FAANG position.
  - scarface_74 2y ago
    Out of curiosity, I looked at how much your company pays. https://www.ycombinator.com/companies/omniai/jobs https://www.ycombinator.com/companies/omniai/jobs This doesn’t come close to what a mid level enterprise CRUD developer could make in Atlanta.
    - themanmaran 2y ago
      I'm from Atlanta, and $175k is pretty square with ATL ranges[1]. It's not a Google salary, but startup comp is always more equity weighted. [1] https://www.levels.fyi/t/software-engineer/locations/atlanta-area https://www.levels.fyi/t/software-engineer/locations/atlanta...
      - scarface_74 2y ago
        My bad, looking at the page from mobile it didn’t look like a salary range. I misinterpreted it. I just saw $125K. That being said, I would hope that position is remote for $175K. I would have jumped on it when I was a pure software developer working in Atlanta. But working locally in SF for that amount?
        what 2y ago
        If you can’t live comfortably on 175k in SF, you have a spending problem.
WD-42 2y ago
The irony of an AI company’s hiring woes due to too much AI.
- themanmaran 2y ago
  Hey we process unstructured data so we're trying not to make things worse :')
no_wizard 2y ago
Interesting to me that 'working with AI' is using OpenAI's API. Thats really working with someone elses AI, no? FWIW I have used it quite a bit, but its not really the same thing as developing AI (nevermind my usual rant that we shouldn't call any of this AI, but I digress)
- lytefm 2y ago
  I'd be fine calling myself a "AI Engineer" for having worked with TensorFlow, PyTorch and scikit-learn and having experience in running AI Models in production. Nowadays, making requests to the API of a LLM provider is what makes an engineer a good fit for working at an AI company? Really?
AznHisoka 2y ago
It begs the question: why not just post your job in your website and nowhere else (ie LinkedIn/Indeed)? That reduces the spam AND gives you a link to point people to. I hear these complaints all the time from companies. To me if you’re leaving your jobs up at aggregators like LinkedIn, you have no right to complain (not singling you out btw, just a general thought)
- themanmaran 2y ago
  Because maybe we'll get a good candidate. The cost is virtually free (30 minutes a day clearing an inbox). And once a week we get a decent application.
  - AznHisoka 2y ago
    30 mins a day isn’t virtually free and last I checked it costs $$$ to post jobs on Linkedin
synicalx 2y ago
> After three years of hearing about AI every day if you've never tested a single inference API then why are you applying to an AI startup. Devil's advocate; why would I have made an API request if my employer has never used that service? Maybe that lack of interest on their part is why I'm trying to leave and get a job in a field that's of interest to me.
tennisflyi 2y ago
> 1. AI Spam. I categorized the inbound we got the other day from a job post. Out of 172 daily applicants, we got 22 that looked reasonably like a person, and 150 that were primarily AI generated messages. Which are pretty easy to spot because they're 500 words of tech jargon and rehashing the job description. Is that not the ideal answer? Those that get to move forward are the one's that just happen to write a message that hits a non-descriptive sweet spot? A fucking Magic 8 Ball
drillsteps5 2y ago
> 1. AI Spam. I categorized the inbound we got the other day from a job post. Out of 172 daily applicants, we got 22 that looked reasonably like a person, and 150 that were primarily AI generated messages. Which are pretty easy to spot because they're 500 words of tech jargon and rehashing the job description. This is what it takes to get through the filters/recruiter search. In majority of the organizations the first line of defense is the recruiter who has a limited knowledge of the job responsibilities, technologies involved, etc. They employ various search techniques in their ATS (lately enhanced by the various LLM tools) and whatever resumes come up in the keyword search will end up in the pile presented to the hiring manager. This is especially true now when volume of resumes is in hundreds, you can't just go through all of them manually. So the only way to get through the first stage is to create resumes to fit the LLM/keyword search... EDIT: I've been seeing more and more of these disclaimers when I submit an application: "We use Machine Learning for an initial comparison of resumes against the education, experience, and skills requirements of the job description." Check here to out out of this analysis. So candidates tweak resumes to match the requirements (likely) with the same LLM that will be checking the resume match to the same requirements. Why is this surprising?

nedwin 2y ago

Adding to this: we mostly focus on outbound, mostly using WorkAtAStartup or LinkedIn as a source.

rors 2y ago

I have a PhD from one of the best universities in the world in Machine Translation, I was training feed forward networks in 2014, built models in TensorFlow in 2016, founded a generative AI startup in 2019 and signed deals with huge consumer goods companies, debugged distributed training jobs on a cluster with thousands of A100s, and wrote PyTorch training pipelines for a financial ML trading signal that made money. Yet, I am unqualified to join an AI startup because I’ve never made an API request to OpenAI

yapyap 2y ago
Make ur first API request to OpenAI today and who knows, rooting for ya!
mikhael28 2y ago
This is a bad faith comment - if you had all that in your resume, they wouldnt disqualify you for not ‘sending an api request’.
beambot 2y ago
The attitude underlying this comment likely makes you a bad culture fit...
sshine 2y ago
Having a PhD vs. making that API call might ironically suggest you don’t emanate hands-on product thinking. I know PhDs in data science and language technology who are not startup material because they’re too uncomfortable outside of MatLab/Jupyter/R.

doctorpangloss 2y ago

The unspoken problem is that most people want way too much fucking money.

themanmaran 2y ago
So far we haven't had a candidate pass because of salary yet. Typically candidates looking for that stable $400k salary never even talk with startups, because they know that's only a FAANG position.
- scarface_74 2y ago
  Out of curiosity, I looked at how much your company pays. https://www.ycombinator.com/companies/omniai/jobs https://www.ycombinator.com/companies/omniai/jobs This doesn’t come close to what a mid level enterprise CRUD developer could make in Atlanta.
  - themanmaran 2y ago
    I'm from Atlanta, and $175k is pretty square with ATL ranges[1]. It's not a Google salary, but startup comp is always more equity weighted. [1] https://www.levels.fyi/t/software-engineer/locations/atlanta-area https://www.levels.fyi/t/software-engineer/locations/atlanta...
    - scarface_74 2y ago
      My bad, looking at the page from mobile it didn’t look like a salary range. I misinterpreted it. I just saw $125K. That being said, I would hope that position is remote for $175K. I would have jumped on it when I was a pure software developer working in Atlanta. But working locally in SF for that amount?
      - what 2y ago
        If you can’t live comfortably on 175k in SF, you have a spending problem.

WD-42 2y ago

The irony of an AI company’s hiring woes due to too much AI.

themanmaran 2y ago
Hey we process unstructured data so we're trying not to make things worse :')

no_wizard 2y ago

Interesting to me that 'working with AI' is using OpenAI's API. Thats really working with someone elses AI, no? FWIW I have used it quite a bit, but its not really the same thing as developing AI (nevermind my usual rant that we shouldn't call any of this AI, but I digress)

lytefm 2y ago
I'd be fine calling myself a "AI Engineer" for having worked with TensorFlow, PyTorch and scikit-learn and having experience in running AI Models in production. Nowadays, making requests to the API of a LLM provider is what makes an engineer a good fit for working at an AI company? Really?

AznHisoka 2y ago

It begs the question: why not just post your job in your website and nowhere else (ie LinkedIn/Indeed)? That reduces the spam AND gives you a link to point people to. I hear these complaints all the time from companies. To me if you’re leaving your jobs up at aggregators like LinkedIn, you have no right to complain (not singling you out btw, just a general thought)

themanmaran 2y ago
Because maybe we'll get a good candidate. The cost is virtually free (30 minutes a day clearing an inbox). And once a week we get a decent application.
- AznHisoka 2y ago
  30 mins a day isn’t virtually free and last I checked it costs $$$ to post jobs on Linkedin

synicalx 2y ago

> After three years of hearing about AI every day if you've never tested a single inference API then why are you applying to an AI startup. Devil's advocate; why would I have made an API request if my employer has never used that service? Maybe that lack of interest on their part is why I'm trying to leave and get a job in a field that's of interest to me.

tennisflyi 2y ago

drillsteps5 2y ago

> 1. AI Spam. I categorized the inbound we got the other day from a job post. Out of 172 daily applicants, we got 22 that looked reasonably like a person, and 150 that were primarily AI generated messages. Which are pretty easy to spot because they're 500 words of tech jargon and rehashing the job description. This is what it takes to get through the filters/recruiter search. In majority of the organizations the first line of defense is the recruiter who has a limited knowledge of the job responsibilities, technologies involved, etc. They employ various search techniques in their ATS (lately enhanced by the various LLM tools) and whatever resumes come up in the keyword search will end up in the pile presented to the hiring manager. This is especially true now when volume of resumes is in hundreds, you can't just go through all of them manually. So the only way to get through the first stage is to create resumes to fit the LLM/keyword search... EDIT: I've been seeing more and more of these disclaimers when I submit an application: "We use Machine Learning for an initial comparison of resumes against the education, experience, and skills requirements of the job description." Check here to out out of this analysis. So candidates tweak resumes to match the requirements (likely) with the same LLM that will be checking the resume match to the same requirements. Why is this surprising?