RTMath is a collection of in-house built .Net and Java libraries for math, statistics, time-series analysis, financial computations, numerical optimizations, and Machine Learning. RTMath components are specially designed to run online computations within strict latency SLAs of high-frequency trading bots.
We have managed to achieve superior results due to efficient integrations with best-in-class native code libraries (like Intel MKL, MKL PARDISO, LibSVM, SharkML), continuous scientific papers, and competitors' offerings research, as well as using various code-optimizations, including designing special memory-efficient interfaces for stream processing.
Unlike most modern technologies, targeting large-scale parallel computations on clouds and big clusters, RTMath libraries show the best results on single-node environments, as they do not waste time on scheduling MapReduce experiments.
RTMath was originally designed as a part of QuantOffice product and to tackle use cases in the financial domain. Today, RTMath is a sophisticated standalone product expanding way beyond the original scope of tasks.
- Core Technologies: Native C, C++, Java, .Net,Python
- Auxiliary Technologies: Cake and Gradle, Nunit, Java, JMH, BenchmarkDotNet, Intel MKL, IPP, BLAS, LAPACK, PARDISO, OpenCV, Shark-ML, LibSCM
RTMath is developed and supported by the Senior Data Scientist Andrei Davydau