Supported and optimized the existing CCFA application (designed for loan risks calculations and written on C/C++). Detected and fixed memory leaks and bugs, implemented new features. Accelerated execution (by 3-4 times) and reduced memory usage (by 30-35%) for the latest production versions. Adapted codes for execution on IBM Minsky servers (clusters of Power8/9 CPU and Nvidia GPU processors). Drastically (up to 5-10 times) improved performance by parallelizing computations and exploiting specifics of target hardware and CUDA platform. Converted applications written on R and MATLAB to Python with subsequent optimization.

Designed, developed, and supported high-performance tools for moving (including extensive ETL – Extract, Transformation and Load and validation) of big data between different data appliances (MS SQL Server, Oracle, Netezza, and Greenplum) and / or legacy file systems (local, remote, shared, and distributed) using Python, Perl, shell, C/C++, Java and SQL as implementation languages for networks of Sparc/Intel servers with Solaris/Linux and Windows operating systems (OS). Performed several legacy data migrations (hundreds of terabytes) from old to new, more powerful data centers. Implemented framework for gradual transition of all data processing (hundreds of extract, transform, load (ETL) jobs and analytical programs) from traditional data appliances to Cloud, enabling continued normal business operation despite significant increase of ingested data volume. Fixed bugs and added new features for operational tools. Monitored, analyzed and improved performance, achieving up to 3 – 10 times acceleration for critical data moves. Provided support of high-performance data archiving and retrieval, making possible re-processing of old data. Facilitated rapid delivery of data for daily processing, back- and re-processing, at twice previous data throughput rate (comparing to earlier used tools including the IBM’s DataStage). Supported legacy framework for FINRA daily data ingestion by detecting and fixing bugs in code and data, adding new features and ETL jobs, generating reports, and collecting statistics.

Developed and supported Intel C/C++ and Fortran compilers for multiple hardware platforms and OS. Fixed crucial bugs reported by customers. Provided code generation in and for Mac OS X environment. Improved use of memory cache for Intel x86 processors and increased generated code performance, up to 40% for some benchmarks. Developed critical components of advanced optimization scheme.

Developed and supported Sun compilers for multiple hardware platforms and OS. Ported Sun C/C++ and Fortran compilers to Intel/AMD platforms with Solaris/Linux OS. Created core complier components (parts of front-end, code generator and optimizer). Invented and improved advanced optimization techniques (performance improvement). Supported existing compiler versions (fixing bugs and adding new features). Extended test bases for Sun products. Contributed to early releases of Java (JDK 1.1).

Project leader in development of Prolog language IDE for multi-processor computer systems