The MAU Accelerator achieves both deterministic low tail latency and high throughput, without trading off one against the other. This enables the use of higher quality models in a given latency and significant savings in infrastructure costs and energy consumption. The MAU Accelerator runs on the Alveo U250 and supports an industry-standard ONNX-based development flow enabling neural network models developed using standard frameworks such as TensorFlow, PyTorch or MXnet to be deployed easily on it. Designed for applications including: Speech Transcription, Natural Language Processing, Speech Synthesis, Time Series Analysis, Payment & Trading Fraud Detection, Recommendation Systems