Myrtle.AI is a Cambridge software company producing, accelerated neural network solutions addressing the inferencing demands of enterprise data owners.Myrtle.AI produces efficient, state-of-the-art, SaS and IP solutions delivering CNN, RNN and LSTM AI algorithms for a wide range of re-configurable, FPGA technology. We focus on accelerating data center and cloud workloads, where low power FPGA silicon is appearing at significant scale. We provide a powerful alternative to current GPU and CPU data center solutions, saving both time and energy. Our technology addresses data center AI applications, such as machine translation, transcription and speech synthesis, where existing hardware solutions under-perform. For recurrent neural networks, we have developed unique, fast silicon circuits, which at batch size one, with unstructured sparsity, enable low latency inference at production scale. Our proprietary sparse matrix IP accelerates core neural network functions, in one case delivering a 42 x performance improvement over specific DeepBench deep learning hardware. The flexibility of our software and FPGA architecture designs, allows us to overcome the hardware constraints of GPU/TPU architectures and to address a wide class of algorithmic approaches. Autoregressive models for audio and iterative inference for video, for instance, can also be output as efficient hardware designs. At the heart of our in-house pipeline, we compile neural network algorithms direct to FPGA VHDL designs, so we can easily adapt this to output deep learning designs for alternative silicon devices, such as ASICs, SoCs and Edge Servers, delivering the same efficient, low latency AI solutions for a wide range of inference applications, outside the data center.