Live data from Hacker News

Show HN: I made CV screening, where each CV battles for Elo score

vkncy.com

1–3 of 3 posts

Show HN: I made CV screening, where each CV battles for Elo score

#1
Hey HN,

I’ve been on both sides of hiring — as a job seeker and as someone trying to build a team. And honestly, the current process is messy. Recruiters drown in resumes. Great candidates get lost. And keyword-based filtering doesn’t really work — especially for technical roles.

So I started building vkncy, with a big vision: an AI agent that can handle everything from screening to interview scheduling. Right now, I’ve shipped the first part — intelligent CV ranking, powered by a dynamic Elo system.

Here’s how it works:

1. Every new CV is parsed and given an initial Elo score by AI.

2. Existing applicants are split into three Elo tiers: high, middle, and low.

3. From each tier, we randomly select one CV (3 total).

4. The new applicant is compared to each of those in a “battle,” where an AI model evaluates and picks a winner.

5. Based on the results, we update the Elo scores using the standard formula.

This happens for every new applicant — which creates a constantly evolving, self-adjusting ranking of candidates for that specific job.

Why this matters: instead of relying on keyword matching or rigid filters, this system adapts. It introduces relative performance, lets different types of candidates shine in their own context, and reduces bias from manual screening.

Right now, you can: - Create a job post - See how the system ranks candidates over time

Next steps:

- Adding depth control: so recruiters can define how many comparisons to make - Building toward an AI agent that also handles voice screening and interview scheduling

Would love any thoughts, feedback, or pushback.

Show HN: I made CV screening, where each CV battles for Elo score
vkncy.com

Re: Show HN: I made CV screening, where each CV battles for Elo score

#3

Very interesting idea. Maybe you can add human in the loop ranking for CVs, which can be used to further tune the AI.

I’ve actually been thinking about something similar! One idea is to add a Tinder-style review flow right after a job post is created — where the hiring team can like/dislike a few AI-generated example CVs. That feedback would help tailor the model to the team’s actual preferences.

Another direction is introducing a “hidden preference” layer. Right now, the AI ranks candidates based solely on the job post — which applicants can see and potentially optimize for. Hidden preferences could let the model learn from team behavior without exposing that data, making evaluations more accurate.