The requirements ?
First up you don't let anyone in without a suitable maths and programming filter.
The maths would be quite similar to a generic quant, but with less time spent on proofs, more on visualising and comprehension.
For the CQF, I worked out a core of
C++, which included patterns, optimisation, threads, Excel addins, Boost, STL, memory management and perhaps most importantly
debugging.
Numerical methods: MC, MCMC, FDM, FEM, etc.
Optimisation methods for portfolios, risk/return Kelly criteria et al.
But a QD also needs a basic hit of Java, VBA and Matlab. Elementary SQL,XML and Perl.
Sucking data out of Bloomberg and Reuters, sockets and FIX.
The way I'd like to do it is a boot camp to get people to a common level of maths and finance, and fill in some holes in programming.
The right way of doing it, represent I think one of the reasons no university does it.
You should emulate the normal cycle of QD work in a bank.
Work through the business rationale and maths of a model then implement it.
Then do it again with a trickier model.
Repeat until you can price a variety of options, and hedge positions in them.