Signaloid Compute Engine Machine Image

Signaloid provides computing platforms for dramatically reducing the runtime and infrastructure requirements of many critical financial and engineering workloads. The target use cases fall into two broad classes: (1) the many real-world applications that today solve partial differential equations (PDEs) by converting them to stochastic differential equations and then solving those SDEs using Monte Carlo methods and (2) the many commercially-important computing workloads that require computations across different input scenarios, which again today often use Monte Carlo methods to build a statistically-accurate approximation across the scenarios of interest. Across these two broad classes, Signaloid's UxHw technology provides a computationally more efficient method of achieving the same results, without the run-to-run variability inherent in Monte Carlo methods.

Signaloid's UxHw technology gives your existing C/C++ applications the abstraction of running over a virtual processor whose floating-point input data can be annotated with continuous probability distributions and whose floating-point instructions performs arithmetic on these distributions, detecting correlations as they occur. With only minor modifications to existing Monte-Carlo-type kernels, achieve orders of magnitude speedup on your existing EC2 instances, from quantitative finance workloads such as value at risk (VaR), expected shortfall (ES), Heath-Jarrow-Morton (HJM) swaptions pricing, derivatives pricing, to engineering simulation workloads such as uncertainty quantification (UQ) of engineering models, probabilistic supply-chain planning, radiation transport modeling (e.g., Bethe-Bloch), and uncertainty quantification of pre-trained machine learning models in ONNX format. The Signaloid Computing Engine AMI provides the Signaloid Computing Engine SDK virtualization layer and the associated utilities needed to compile and run C/C++ applications to take advantage of Signaloid's UxHw technology. By simply replacing your default AMI with the Signaloid Compute Engine AMI and recompiling your applications using the Signaloid UxHw SDK, you can achieve speedups on your same EC2 hardware instance types, of up to 430x for VaR (using geometric Brownian motion), up to 580x for HJM, up to 34x for engineering model UQ, and up to 56x for Bethe-Bloch.

Keywords/Phrases: Monte Carlo Simulation, Quantitative Finance, Stochastic Differential Equations, Value at Risk, Expected Shortfall (ES), Heath Jarrow Morton (HJM), Valuation Adjustment (xVA), Probabilistic Supply-Chain Planning.