I am director of graduate admissions for a math Ph.D. program.  My main advice is not to use recommendation letters from your current employment -- they are very unlikely to offer information relevant to your potential in research mathematics, even given that your job is highly quantitative.  Instead, get letters from your undergrad math professors.  You don't say how long you've been out of school -- if you're concerned your undergrad professors don't remember you, I would recommend sitting in on a course or two at a university near you, if the professor doesn't object.  This will both a) give committees a chance to hear from someone who's talked math with you *recently*, and b) give *you* the chance to make sure that spending a lot of time learning math is what you want for the next phase of your life.

The advice to study hard for and do well on the subject GRE is also sound; the GRE has its problems, but it can be a good bolster for an applicant whose qualifications are unconventional in other ways.