I did a genetic algorithm assignment a couple of semesters ago for my Masters and I literally took the Wikipedia article and used it as a "recipe". The implementation is really not that hard for genetic algorithms. The hardest part is coming up with a fitness function that rates how good the solution is.
What you need to do is:
1. Define a genetic mapping for the solution. This can be either a string of bits or an array of some kind of integer. Your program will be able to decode the mapping into a solution for whatever problem you're trying to solve. In my case, it was the "n-queens problem", and I had an array of integers where each element was a row and the value was a column.
2. Define a fitness function. In my case, I would count the number of queens that could capture other queens. The lower the score the higher the fitness.
For the specifics on the algorithm, see Wikipedia: