Collected US 2024 tech job postings from Indeed and embedded them with Open AI text embedding large. Reduced dimensionality and clustered via UMAP and HDBSCAN. Topic modeled with Open AI chat API. Visualized with DataMapPlot. Github pages has full interactive map. I also have real-time insights into tech job postings on my site [hazon.fyi](http://hazon.fyi)
Software development is a little vague.
Like to know how much of that is for mobile apps and gaming
This is really hard to grok.
Cybersecurity is its own little island 🏝
aaaaaand 90% of them are fake
How have you got HR and not Sales? There’s no way there are no sales jobs in tech paying under $250k.
Anyone else see the outline of Australia?
are these supposed to be on a map? cause i have no idea where on a map any of these lie
How do you get around Indeeds Cloudflare, I manually copy the html from like 100+ pages every week :’)
I’m not very good at making my scrapers look human, I should learn selenium…
Also you said you used UMAP and then HDBSCAN, did you use one after the other to get wider groups, then subgroups?
9 comments
Collected US 2024 tech job postings from Indeed and embedded them with Open AI text embedding large. Reduced dimensionality and clustered via UMAP and HDBSCAN. Topic modeled with Open AI chat API. Visualized with DataMapPlot. Github pages has full interactive map. I also have real-time insights into tech job postings on my site [hazon.fyi](http://hazon.fyi)
Software development is a little vague.
Like to know how much of that is for mobile apps and gaming
This is really hard to grok.
Cybersecurity is its own little island 🏝
aaaaaand 90% of them are fake
How have you got HR and not Sales? There’s no way there are no sales jobs in tech paying under $250k.
Anyone else see the outline of Australia?
are these supposed to be on a map? cause i have no idea where on a map any of these lie
How do you get around Indeeds Cloudflare, I manually copy the html from like 100+ pages every week :’)
I’m not very good at making my scrapers look human, I should learn selenium…
Also you said you used UMAP and then HDBSCAN, did you use one after the other to get wider groups, then subgroups?