Summit Peak Technologies, LLC (SPT) is dedicated to computer technology transfer in all fields.

 

Since 2004, SPT has conducted research and development for slope stability, rockfall mitigation, ground imaging, and deep foundations. SPT has made significant breakthroughs for increasing efficiency, boosting productivity, and bringing realism to geotechnical engineering with SNAP-2 for FHWA (Federal Highway Administration).

​

In 2014, SPT became the first company to utilize the GPU for high speed parallel super computer computation, implementing the Colorado Rockfall Simulation Program (CRSP-3D). 

 

SPT wrote GpuScript, the first GPU language with full debugging support, allowing entire programs to be ported to a GPU using Object-Oriented Programming (OOP) design. GpuScript is a paradigm shift in programming, increasing programmer productivity and program computation efficiency by orders of magnitude.

​

SPT developed Geometric Empirical Modeling (GEM), the solution for neural networks. The full neural network structure with thousands of hidden layers, nodes, connection links, weights, and offsets are instantaneously built and computed for any training set in a single GPU call, in milliseconds. No trial and error to find the right structure, no learning rate, no iterating over each example in the training set over and over, no local minima, no over-fitting or under-fitting, no errors, no problems with correlated inputs, high dimensionality, non-linearity, clustering, extraneous inputs, missing or unknown data, outliers, jitter, scatter, or data distribution. GEM instantly gives mathematically optimal solutions with unprecedented interpolation and extrapolation. GEM has instantaneous thinking, or the ability to determine inputs given constraints on other inputs and outputs. GEM is a paradigm shift in AI, eliminating the need for AI experts and AI super-computing, including AI factories, GPU super-clusters, TPUs, and MXUs. 

​

An open source version of GpuScript is available on GitHub.