So there are problems with hiring engineers that no one is working on. Those problems are:
- The engineers who create bad resumes
- The hiring managers who write bad job descriptions
- The entire HR process, which has no feedback, and relies on matching bad job descriptions with bad resumes.
Perhaps you’re thinking, that’s pretty much the whole system. Yes, the system is broken. This is because of the following:
-
The best engineers often have the worst resumes, relying on the jargon from their old job.
-
Hiring Managers often tend to draft Job requirements that are what I call “Lawyer turned Pastry Chef Unicorns”, also relying on internal jargon.
- Resumes and Job Descriptions are a lossy medium of information exchange, making automation harder.
Ah, but X can fix that! Or rather, x.ai can fix that. Job hunting isn’t scaling because it relies on a overly manual screening process.
First, some philosophy on hiring: Hiring Athletes instead of Bigfoot
So how can X.ai help? Essentially by using their existing AI tech to upgrade all the pieces along the way at a bootstrap process with the long-term goal to replace the lossy traditional resume exchange method with a complete knowledge system.
It can help Hiring Managers write better Job Descriptions. This should be an iterative, intelligent process. One of my best hires told me “I don’t know Java”. I told him, “don’t you know C++?”. He said yes. I said, “I think you’ll figure it out.”. He did. This was Joel Goldberger, one of the engineers who really did invent the internet. Specifically, he invented email. (He wrote sendmail for tops20, and the POP2 protocol, before his nemesis added 2 commands to the RFC, and instead of submitting it as an update, he submitted it as a replacement RFC POP3.) His nemesis went on to invent SNMP the Simple Net Management Protocol that is anything but simple. His nemesis invented spam too! When Joel emailed him and said “I don’t think we should accept email from just anyone, and his nemesis responded with ‘you’re tops20 and I’m unix so I win’”. So not only did Joel invent email, he tried to prevent Spam back in the day. Joel has that email squirreled away somewhere if we ever we want to string that guy up… SNMP and Spam being some of the greatest evils ever inflicted on the internet. But I digress. Basically the point is that Joel was the shit, I knew he was the shit, but he didn’t know that. So I had to convince him to work for me. (He hates his new manager though, so if you want to hire him, and you should, reach out to me.)
Most jobs are filled by the Hiring Manager hiring someone he or she knows which should tell us, the system is broken.
As a Hiring Manager, it’s natural to think, “well, I need a lawyer, and I’d like a croissant, so I really need a Lawyer who became a Pastry Chef”. Now, being a lawyer is one of the worst things in the world, so no doubt there are 3-5 people who fit that description, they were sitting in their fancy office doing lawyer shit (or a shitty office), and they realized “I don’t want to do this anymore”. But you won’t be able to find them or hire them, they’re quite happy going to their strip mall bakery every morning, it’s so much better than being a lawyer. Or for a less silly example, you might use BitBucket instead of Github at your company, so you put that in the Job Description. So Hiring Managers need guidance in writing their JD. Also, technology moves fast, I made a list the other day of all the technologies I liked and what version I was last using and what version was new, and it was pretty shocking. Being an engineer means keeping continually learning. As I said, you need to hire athletes, not the perfect kitchen utensil. Attitude is everything, you can train people on your special flavor of tech.
On the employee side, writing a resume is marketing. That is a different skill than engineering. So the best engineers suck at that. They put down terms that are specific to their company like say:
Accelerated the Top Hat system 2x.
Ok, he accelerated something, but what? As a hiring manager, I don’t know what to think of that. Main feedback I give people when reviewing resumes is to describe the thing, instead of using the company jargon for the thing. This is where a LLM approach could extract the actual content by asking questions.
What does the Top Hat system do?
It’s the system that processes all of our major customers, the “Top Hats”.
Ok, how did you accelerate it?
I built a custom profiling system, used it to find bottlenecks, fixed those.
Ok, is it fair to say “Built custom profiling system to accelerate the processing system for our major customers 2x”?
Yes.
Ok, adjusting resume. What was the business effect of that?
It saved us $2M in cloud expenses the first year.
Ok, is it fair to say “Built custom profiling system to accelerate the processing system for our major customers 2 x, saving $2M in cloud costs.”
Yes.
Ok, adjusting resume. What did the Spats system do?.
Ah well, that’s a story. See…
Another issue for engineering job seekers is that most of us have several tech skills, and if the exchange mechanism for hiring information is a broken job description talking to a dumbed-down resume, properly applying to a job requires customizing your resume for each job to pass the six-second man-from-mars test. (That’s not about Elon, it’s a UI term for coming at something fresh, famously recruiters look at a resume for 6 seconds. That’s about all they can do given the resume volume they have to deal with.)
Further, the AI can refine skills and languages into related clusters, and group them accordingly.
If you’re looking for Java, C++ will probably do.
If you’re looking for J2EE, Hibernate will probably do. BitBucket is a more enterprise friendly version of GitHub, etc.
The way we solve this now is HR has special people called technicalrecruiters who know enough about engineering to translate between the JD and the resumes. (Though I doubt they know that BitBucket and Github are essentially equivalent.) If only they weren’t completely overwhelmed by the volume of resume submissions.
So ultimately, we have a Job Description, which is more noise than signal, and we have a Resume, where much of the signal has been limited, 1 page, etc.. Data has been lost at each stage.
So we have the steps needed here to fix this:
- The AI will work with Hiring Managers to turn a Job Description into a Job Model. Output is a better job description. Perhaps HMs can even get feedback from the existing set of candidates, so they know to ask for git experience instead of BitBucket.
- The AI will work with candidates to refine their resume into a Person Model. This will encapsulate all of their experience, more of a CV than a 1 page resume.
- As an incentive, the x.ai jobs system will allow candidates to create a resume for free, a deep resume, and candidates can manually get a tailored resume if they provide a job description.
Then, instead of relying on this lossy resume process to exchange information, the AI will work to find matches between the Person Models and the Job Model.
X.com could bootstrap market share by starting with a nominal $10 fee to avoid bots, an interactive resume builder for people. That will loop in the job hunters, which are 10x the volume. Same thing with the hiring managers, loop them in to write Job Descriptions. This is to get to the point that the network effect takes over, and it’s natural for X to have both sides. The companies hiring will then pay per hire for creating a Job Model and surfacing candidates to them because in hiring, its the companies that have the money.