Salary explorer based on US job postings: https://corvi.careers/salary-explorer/

I index all of them into a standard schema using various techniques including using LLMs for some part of it.

I am using openai small embedding model with custom compression. It is super cheap. You can read more at https://corvi.careers/blog/vector-search-embedding-compressi...

Write QPS isn't what you want to measure, you should look at index fragmentation, dead tuple usage etc etc. All of those generally go in directions that you don't want.

I am building corvi.careers, its a job search engine not social network tho

http://corvi.careers - job search with features like resume match score etc

The problem is way worse if you update rows, if you stick with insert and delete you can get quite far.

In Postgres, an UPDATE is effectively an INSERT + DELETE [0].

0: https://www.cybertec-postgresql.com/en/is-update-the-same-as...

try https://corvi.careers I been building it purely as job search platform, has good coverage of startups and public companies but I’d still recommend to use LI for network tho

Few observations from analysis of 100K+ open job postings (from a job search site I am working on) in swe/ai/sre/devops. (ex: China).

Nothing too surprising but am keeping an eye on - Applied AI (Mention of Claude etc as one of job requirements) - Hiring trends around entry levels. (Most openings still heavily concentrated at mid/senior bands)

From what I could see, big retailers have a lot of "evergreen" openings which makes sense as they can have multiple locations and there is a lot of churn. And there are obvious outlier sub-categories like warehouse workers etc which have median times <7d, I didn't break it down in the blog as it's too much data to present. But other than that, I don't have enough search data to draw meaningful conclusions. (say around supply/demand)