The Job Seeker project so far

It's been an eventful week! We started a community project based on the Job Seeker project pitch on our Miro Board. I've developed one or two not-so-small projects in the past, but never have I had to organise a GitHub repository with pull requests. I'm doing so now, making all the errors one can imagine.

Here's the progress we made so far:

We made a web scraper to scrape job postings from linkedin
We prompted the LLM to generate skill key words for every posting

Not too shabby for a first time project! There's still lots to cover, mind you:

Prompt the LLM to cluster the key words, as they often are too detailed
prompt the LLM to analyse a resume for missing keywords from the cluster list

And then, a UI. Probably a web page. I imagine something like:

Have a set of key words for job titles we already scraped
If the user happens to ask for an existing job title, the LLM immediately asks for their resume
Otherwise, we start scraping (long process)
Then we generate the keywords (long process)
Then we generate the clusters (long process)
Due to the time involved for steps 3-5, this cannot be a real time process. I suspect we will have to send an email to the user. Oh no! Email servers? The road to hell is paved with good intentions...
Now the LLM asks for the user's resume
Magic happens.
Profit: the user hears the LLM's recommendations about which keywords to add to their resume.

Want to help out? Come to our Slack channel and offer your services! ;)

Thank you

@Brandon Phillips

@Ana Crosatto Thomsen

@Olu Akin

, Slav Petkovic (where's he gone?),

@Sapnil Patel

for making this fun, inspiring and who knows perhaps successful 🎉

11 comments