Show HN: I used an LLM to parse HN Who's Hiring comments to structured job data (opens in new tab)

(hnjobsai.vercel.app)

16 pointstekmaven2y ago8 comments

Hello! I used an LLM to extract structured data from each comment on the “Ask HN: Who is Hiring? (January 2024)” post, available at: https://news.ycombinator.com/item?id=38842977. This created a dataset of 400+ jobs with data such as title, company name, location, is job remote, salary data, and the URL/email to contact. For the UI, a NextJS static site was created to filter and view the structured data.

8 comments

alecsm2y ago

Funny that only 51 out of 475 jobs show the salary while tech people are very vocal against not showing it.

quickthrower22y ago

Good stuff. Remote is not a boolean though. It would be good to specify:

* Which countries can apply for the role (or is it Global)

* Which timezones can apply for the role (or is it Global)

If not specified you could assume you need to be a citizen of their country and in their time zone +/- 3 hours.

I think this is where an LLM can shine, it may be able to make good guesses about this stuff.

helsinkiandrew2y ago

Nice, a time/date on each comment might be useful, so you can find the jobs posted an hour ago rather than 5 days.

tekmavenOP2y ago

Thanks for the suggestion. I added the time posted as the first piece of displayed metadata. Also, I updated the sort order to display the newest jobs first.

LorenzoGood2y ago

Awesome!

Small suggestion, move the "Link to original comment" button, to the HN comment drop-down.

tekmavenOP2y ago

Thanks for the feedback! I've moved it below the original comment now. I made some small adjustments to the original comment display as well.

chrissnell2y ago

Can we get a "Kubernetes" added to the technologies filter?

niemyjski2y ago

Pretty awesome!!

j / k navigate · click thread line to collapse

Show HN: I used an LLM to parse HN Who's Hiring comments to structured job data (opens in new tab)

(hnjobsai.vercel.app)

16 pointstekmaven2y ago8 comments

8 comments

alecsm2y ago

Funny that only 51 out of 475 jobs show the salary while tech people are very vocal against not showing it.

quickthrower22y ago

Good stuff. Remote is not a boolean though. It would be good to specify:

* Which countries can apply for the role (or is it Global)

* Which timezones can apply for the role (or is it Global)

If not specified you could assume you need to be a citizen of their country and in their time zone +/- 3 hours.

I think this is where an LLM can shine, it may be able to make good guesses about this stuff.

helsinkiandrew2y ago

Nice, a time/date on each comment might be useful, so you can find the jobs posted an hour ago rather than 5 days.

tekmavenOP2y ago

Thanks for the suggestion. I added the time posted as the first piece of displayed metadata. Also, I updated the sort order to display the newest jobs first.

LorenzoGood2y ago

Awesome!

Small suggestion, move the "Link to original comment" button, to the HN comment drop-down.

tekmavenOP2y ago

Thanks for the feedback! I've moved it below the original comment now. I made some small adjustments to the original comment display as well.

chrissnell2y ago

Can we get a "Kubernetes" added to the technologies filter?

niemyjski2y ago

Pretty awesome!!

j / k navigate · click thread line to collapse