FFT is short for fast Fourier transform, any of a set of algorithms for quickly computing the discrete Fourier transform (DFT).
The FFT finds a lot of application in data analysis, particularly time-series and image data, and particularly when the data has a periodic nature, or at least a periodic component. The FFT also finds application in digital filtering. There are many FFT algorithms; they all calculate the Discrete Fourier Transform in O(n log n) operations, while the naive DFT implementation is O(n^2).
Mathematically, the Fourier Transform fits a set of sinusoids to the input data - revealing relative strengths of periodic components of the signal. The fit is optimal in a least-squared error sense. In the case of the Discrete Fourier Transform, the sinusoids are periodically related.
Related topics include DFT, signal processing, convolution, and window functions.
More information on FFT can be found in the Wikipedia article on FFT.