Thread by @sanketsaurav

If resume parsing and matching it for job fit actually worked, we'd have already solved technical hiring. It doesn't, and anyone claiming otherwise is either a charlatan or hasn't launched their product yet to real people recruiting engineers.

https://twitter.com/banglani/status/1268919831962714114

In the absence of an objective source of truth about one's skills (which resumes aren't, and neither is your LinkedIn profile), any sort of matching to jobs is a farce. Most companies trying to do this believe they can solve it using NLP-based extraction and string matching.

But one can write ANYTHING on their resume. And most companies DO NOT know how to write a job description that actually mirrors what the job actually needs or entails.

So all "data science" happens on junk datasets. Just thinking from first principles.

It's easy to label resume parsing as AI and intelligence and peddle it to recruiting teams. It's hard to actually show any results. Don't believe me? Look at the graveyard of Indian companies who've tried to solve tech hiring.

Solution?

1. You need objective data about people's skills that can be trusted. (≠resume)

2. You need to quantitatively express what a job role needs, and profile the company's historic data on what skill/experience matrix has worked in the past for that role. (≠JD)

Naukri seems to be irreplaceable because we're yet to see someone crack this solution with a product. Not because the technology doesn't exist, but because the data needed to make this work is either inaccessible (in private sources) or hard to quantify.

Fin.