To do it right, you would need access to large sets of genomes with corresponding medical histories. No one has ever really gotten enough data together to be useful yet.

It is an untouched market though, someone will do it eventually and genomic medicine will suddenly be very useful. I would do it if I had more capital. Also, it would require competing with a public project: http://www.personalgenomes.org/participate.html. I'm not sure how well funded that project actually is, it won't be feasible for at least 5 years.