Hi, I am very familiar with this topics. I have been working on similar problems during my PhD research. I have a strong background on audio processing, music information retrieval, data mining, machine learning.
Did you already trained models to recognize genre/timbre etc..? Or do you need the training phase as well? Do you have a suitable dataset?
Please check that the budget is ok for you. This is the very minimum that I can ask. Please let me know