Taras Rybalchenko's Colleagues at Integrated Technology Laboratory LLC
Taras Rybalchenko's Contact Details
N/A
Ukraine
Integrated Technology Laboratory LLC
Taras Rybalchenko's Company Details
Intela.io (Integrated Technology Laboratory LLC) provides cutting-edge computing platforms to tackle real world problems. With more than 200 specialists and associated teachers/professors across the globe Intela provides wide range of services including- Cloud platform development;- Big Data aaS;- Data Science aaS;- Hyperledger blockchain platforms and networks;- Scientific research;- Education and Educational platforms for UniversitiesCompany interest is directed to educational platforms development and support, including implementation of online education within universities, student performance tracking, social networking provision, student projects promotion. You can learn more at https://www.intela-edu.com/ Another company focus is 'in-house' product development.MathDB (Desktop, Cloud and Embedded for IoT versions) technology empowers Big Data and Machine learning processes by abstracting database interface over time series data, implementing fast minimal-loss streaming queries by exploiting structural features of the stored data. MathDB time series data is stored through a mathematical decomposition based on Fourier and Wavelet modes in a way that is optimized for streaming queries. As a feature unique to MathDB, such queries can be made faster by the selective recombination of the stored decomposition. By only accessing parts of the stored data at a time, MathDB reconstructs the data in a fraction of the IO access time with minimal loss of fidelity. This abstracted database interface has shown promise in data science applications by enhancing many existing tools and workflows, as well as opening the door to new potential innovations. The most direct benefit is to the runtime of streaming IO-bound algorithms (such as training models or performing statistics on large data sets that cannot be stored in memory). Read times have shown to be reduced by factors of 50 and 100 with little or no loss in model predictive capability in real data sets, speeding up ANN training or statistical model fitting by orders of magnitude.