Also, in terms of hallucinations, you might want to consider giving more context to it. So instead of just telling the LLM the domain/company only, scrape their website, get all the javascript and XHR calls it makes, and find any job postings, then pass ALL of that to the LLM. The more data it has, the more it can extract. If you just give it a company website, it's just going to make stuff up. That's my 2 cents if you insist on going the LLM route. Or just tell it to scrape more stuff (which will cost more tho)