Hacker News new | ask | show | jobs
Offer HN: A better "HN Whose Hiring Process"
4 points by hackingla 4893 days ago
HN whose hiring seems like a very inefficient process and I think that two of the products I have created (setatimetotalk.com and cvchk.com) could really improve this process.

I see it working like this

Employer

1) drag and drop resume of ideal candidate

2) view similar candidates

3) pick which to speak to

4) set a time to talk to them

5) get phone call when its time for meeting

subscribe to get email updates of new candidates

Employee

1) drag and drop your resume

2) view jobs you are qualified for

3) pick jobs you like

4) set a time to talk to them

5) get phone call when its time for meeting

subscribe to get email updates of new jobs

The employee part is quite ambitious but I am very proud of the progress I have made.

1) User upload resume

2) I use NLP to map resume items to properties of agent model example: 4 years RoR

3) I then evolve the agent into the ideal candidate for matching jobs (matching == ts vector right now but i will improve)

4) measure the distance that the agent (applicant) needed to evolve to become "ideal" example: ideal = 6y RoR

5) rank and classify based on distance of evolution required

First I want to do jobs then later on doctor -><- patient and lawyer -><- client matches I will have these things ready by the 1st when its time for "HN whose hiring" and will offer for free all I ask is that you help me to make this process better.

2 comments

I think the whole point of HN Who's Hiring is that it's on HN using the message format as the rest of the site. That's like saying that Twitter is very inefficient for news publishing because we should be using Delicious instead... well it's efficient for people on Twitter.
NLP is used extensively in resume screening at the moment and no-one has managed to perfect it. I'd be interested in seeing the results/accuracy of a search against a large (500+) cross section of resumes uploaded.
I think I have had slightly better results than average because I am using ABMs that learn and have social media as precepts. There is a "training" period for each agent where it is trained based on LinkedIn data etc... but I 100% agree with you, I am also interested to see how it works; currently I have not tested more than 100 at a time.
I worry that you may be underestimating the competition. bondadapt.com are a prime example. Recruitment companies pay seven figures annually for their system and they've invested an insane amount of time and money into improving their NLP system and, speaking as someone who's used it extensively in the past, its results are relatively poor when you are dealing with a database of tens of thousands of resumes.

Please don't get me wrong, I'd love to see you crack it and if you manage it be sure to share the results!