Athena provides end-to-end speech processing capabilities including automatic speech recognition (ASR) with hybrid attention/CTC methods, text-to-speech synthesis via FastSpeech/Transformer models, voice activity detection, and key word spotting. Built in TensorFlow >= 2.0.1, it supports multi-GPU distributed training with Horovod, includes a Kaldi-free Python feature extractor, and offers C++ runtime decoding with WFST-based decoding for deployment. The project provides recipes and examples on open-source datasets for various speech tasks.