How We’re Reimagining Candidate Matching with Recruiter-First AI

My espresso cup and I have been in the tech recruiting space for a while now - and while many talk about automation to screen candidates for a job, in practice choosing the right person is far from the classic repetitive, standardized activity to be replaced by AI.

It is instead a very ad-hoc experience, and we wanted to build something that reflects the key traits of recruiting that we were identifying:- Skills matter, yes, but preferences matter more. Two candidates with very similar work experiences can be wildly different fits depending on the company, team, or hiring manager.- Requirements evolve.

You start with a wishlist, then meet a few people and realize… you actually want something else. What if recruiters could choose the requirements they care about most, tweak them any time, and instantly see how candidates are re-ranked? At Pariti, we worked on this idea for months, and we have now launched internally the new candidate recommendation app, which will soon open to the public.

‍

‍

Here's how it works:

We extract what matters from the job description -> We scrape + normalize CVs -> We score candidates with a combo of embeddings + Boolean logic -> Recruiters get a ranked list they can reshape on the fly

‍

‍

On the tech side:

- Vectorization is done using an open-source model- CV parsing is done with proprietary model via API- The real work of art is combining real-time embedding computation, Qdrant, our internal DB, and a fast & interactive frontend.If you’re still reading, you might be as nerdy about hiring as we are

A quick Loom demo is linked below.

‍

Big kudos to the team for building something so powerful and usable.