Job Search Harness
Parallel LLM review of open job postings
I should not apply to most job postings! Even most remote product manager roles aren't right! Even if I manually configure all the filters at some job aggregator site, still - low hit rate!
So 90+% of the JDs I read are a waste of my time. With the right context and structure, we can get LLMs to churn through a lot of that grunt work.
So the goal is context and structure.
We need a process to evaluate companies against the verticals I am interested in. For now, that is an exclusionary filter (e.g. remote friendly. no ecommerce, crypto...)(much of this can be done deterministically). Then for each non-disqualified company we have the LLM evaluate each role for mutual fit, and produce a strucutred priority score (this is the part that wasn't feasible in 2022).
Arichitecture below, but I'm not posting the source on this one publicly because I'm using an undocumented API to source my job listings 🙂.
Harness architecture
unevaluated
evaluations
unevaluated
evaluations
feed the dashboard