Design Shazam

Design an audio recognition system like Shazam. A user holds their phone up to a speaker for 5-10 seconds, the system identifies the song from a catalog of 50M+ tracks, even in noisy environments like bars and concerts. The system must handle 100M+ recognition requests per day with sub-3-second identification latency. Cover audio fingerprinting, spectrogram analysis, noise-robust matching, and large-scale index lookup.