IF done right, 500 can be representative. Although considering the number of jobs out there combined with age groups and gender, I guess you'd need a few thousand carefully selected participants to get valid results.
No need to carefully select participants. If you do random sampling with a large enough sample, the proportion of all subcategories should reflect the population at large.
