HayStax

Like everyone else, I've experienced the fact that when you're applying for work, your resume goes into the black hole of ATS, and oftentimes missing that one specific keyword means that your resume never gets seen. Getting my first real job out of grad school, I realized how lucky I was to get it because my PhD advisor was well-connected and had put my resume in front of people he knew. What was more ironic to me was that I was pretty sure that I had submitted my resume on my own to the group I was eventually hired into 6 months prior.

Now that I'm a bit further along in my career, and have been a hiring manager, I realized that the problem I faced earlier will increase in scale. Every year since 2011, the volume of resumes that any given position has seen has increased. The first time I had to hire someone, it was sufficient that I split the resumes amongst the team, got everyone's top 2-3 choices, and repeated the process until a small number of people could be interviewed. However, the scale of the problem has exploded so that this isn't feasible.

The traditional solution for this has been to use Applicant Tracking Systems (ATS) to be more and more precise in keyword searches so that a larger and larger fraction of applications can be eliminated from contention. However, this is just an arm's race in which people have been tailoring their resumes to fit the exact job description, thus making this a method that has lower and lower utility as time goes on. Other people have leaned into their networks to find candidates that they already know. Furthermore, amongst the people who try to apply without trying to play the system, traditional ATS are terrible at looking for supplementary skills. For instance, for a software developer, I would probably take a Java programmer even if the job description stated C#. A good Java developer generally is able to learn C# relatively quickly because of the similarities in syntax, and it's actually their ability to break down the problem into logical tasks that I'm most interested in.

I still drink the meritocracy Kool-Aid, and believe that it's still possible to hire people fairly rather than those who try to game the system. During my latest hiring experience, we posted a position for a data scientist and prior to blasting it out on traditional job sites, we had more than 500 applications in less than 24 hours. Had we posted it on LinkedIn, Monster, or Indeed I'm sure we could have expected to see 4x that number in the same amount of time, something that would have clearly swamped everyone on my team to review the resumes. I realize that a few of the things that I really need are the following:

• The ability to see if someone has a "nice to have" without ruling them out
• The ability to weight the different "nice to haves" with different numerical scores
• Because many of the applications are from people who are educated aboard, the ability to plug in any gaps in my knowledge about their educational system. For instance, while I know that IIT is India's best university system. Anything besides that is a black box for me, and being able to rank schools directly in the system is important to me. Likewise if you're a Recruiter, holes in your technical knowledge can be filled by the LLM.

So I initially wrote a simple python script to send resumes to an LLM and pull out and score these elements, and I think it worked pretty well for me. HayStax is my attempt at providing this to more hiring managers so that hiring can be fairer and less painful I created this tool.