Development of specialized services

1. Numerical analysis – effective parallel implementation of computer modeling and simulation algorithms of complex dynamic systems (FDM, FEM, FVM, Monte Carlo, Molecular Dynamics), developing modules for common FEA-analysis and visualization systems (OpenFOAM, Paraview).
2. Image processing – massive parallel implementation of batch image processing algorithms (computer vision, analysis of geo, medical and remote sensing images).
3. Data relay and processing services – developing effective fault-tolerant server software with automatic scaling to accept, store and process data from client devices (wearables and mobile applications), working with public (Microsoft Azure) and dedicated clouds.
4. Hardware selection and administration of heterogeneous computing systems – choosing equipment of leading HPC-vendors to meet project requirements, administration of any existing entry-level and mid-level computing systems, adoption and configuring of project-related system software.

Completed projects

1. Solution of the Dirichlet problem for the Poisson equation by the method of successive over relaxation with red and black reordering (RBSOR) on NVIDIA and AMD GPUs.

A library with C and C# APIs allows to perform computations up to 40 times faster compared to regular multicore CPUs.

2. ParaView controlling module for NPO Saturn.

A module to remotely control ParaView visualization within Presentator system has been shown on the customer’s stand during the International Aviation and Space Salon MAKS in Zhukovsky. Results of CFD-analysis of a modern gas turbine engine blade has been demonstrated.

3. High-performance classification of multispectral remote sensing images in Euclid-Mahalanobis metric space on NVIDIA GPUs.

A classification method based on Euclid-Mahalanobis metric space is introduced for remote sensing images. A parallelisation mechanism relies on OpenMP for CPUs and CUDA for NVIDIA GPUs. ASTER images are used for classification speed benchmarks.
classif_eng

 4. Cloud platform for collection and analysis of biomedical data for Kriotekh LTD.

A distributed system comprising  mobile applications (Android and iOS), scalable fault-tolerant cloud platform and web application featuring adaptive interface. Load testing and emergency testing (testing platform resiliency in case of failure of equipment) succeeded for 10,000 patients and 1,000 doctors. Adoption took place on FAP Krasnaya Gorka, Rybinsk district and City Hospital № 6 Rybinsk. The project was mentioned in scientific publications and presented on U-Novos exhibition in Tomsk (2014).