Bayesian stats, Machine Learning, and internet match making

If anyone out there is interested in going forward with it, I have an excellent idea for taking on E-Harmony. E-Harmony uses manifold compression algorithms to match people across 29 dimensions. They ignore a great deal of information, and don’t take into account any Bayesian Stats, Machine Learning. Sorry for being vague, but if you interested it is actually a well fleshed out idea. I have nothing against E-Harmony, so if someone at E-H reads this I and willing to help improve your algorithms..,