As a Technical Lead AI, I am currently leading a small team aimed at developing higher level cloud services targetting AI applications on top of the existing cloud platform. The team is largely acting as an internal customer and consumer of the platform, and driving its development to meet the needs of modern AI.I am leading PoC work with large enterprise AI software vendors and evaluating the ways to deploy it on our cloud platform. My work also involves prototyping, keeping track of… Show more As a Technical Lead AI, I am currently leading a small team aimed at developing higher level cloud services targetting AI applications on top of the existing cloud platform. The team is largely acting as an internal customer and consumer of the platform, and driving its development to meet the needs of modern AI.I am leading PoC work with large enterprise AI software vendors and evaluating the ways to deploy it on our cloud platform. My work also involves prototyping, keeping track of modern AI developments, and providing technical expertise in AI and HPC for business units of the company.As a smaller portion of work, I am still involved into benchmarking and developing a supercomputer experience in a cloud. Show less

As a principal HPC engineer I largely continued working on benchmarking and performance improvements of the cloud platform, as in my previous position. This work included extending the set of benchmarks and working with colleagues on automation and integration to deployment process. I was also responsible for performance benchmarking of our newly developed bare-metal cloud product.I co-led adoption of Infiniband to our existing GPU installation. My area of work included getting familiar… Show more As a principal HPC engineer I largely continued working on benchmarking and performance improvements of the cloud platform, as in my previous position. This work included extending the set of benchmarks and working with colleagues on automation and integration to deployment process. I was also responsible for performance benchmarking of our newly developed bare-metal cloud product.I co-led adoption of Infiniband to our existing GPU installation. My area of work included getting familiar with OFED and UFM, user-facing integration to MPI and GPUDirect RDMA, communications to NVIDIA. I also largely contributed, mostly via R&D and problem solving, to developing Inbiniband operation scenarios, support for multi-tenancy, extending our deployment automation for ConnectX firmware updates and connectivity testing. The resulting Inbiniband tools are currently used in production on scale of a few thousand GPUs.I transferred a lot of GPU knowledge to the support & operations team and together we revamped GPU operating procedures and troubleshooting process. Show less

As an HPC engineer I contributed to various performance- and GPU-related aspects of our cloud platform.My work mainly concerned performance optimization: studying community tools and benchmarks, establishing workflows to run them as internal customer, identifying and fixing performance issues of the platform. The benchmarks used included MLPerf Training and Inference, HPL, OSU Micro Benchmarks, STREAM. As a direct result of my efforts, the team identified and fixed a number of… Show more As an HPC engineer I contributed to various performance- and GPU-related aspects of our cloud platform.My work mainly concerned performance optimization: studying community tools and benchmarks, establishing workflows to run them as internal customer, identifying and fixing performance issues of the platform. The benchmarks used included MLPerf Training and Inference, HPL, OSU Micro Benchmarks, STREAM. As a direct result of my efforts, the team identified and fixed a number of performance issues concerning pinning, NUMA configuration, RAM performance.Another direction of my work involved being an in-house GPU expert and contributing to end user experience via creating docs and OS images with GPU software stack (drivers, Fabric Manager, CUDA, etc.). Show less

As a postdoc in the computational radiation physics group, I am a core developer and maintainer of C++ exascale particle-in-cell code PIConGPU. My work involves implementing and extending core particle-in-cell routines, including state-of-the-art numerics for incident lasers and field absorbers. I take part in preparing and tuning PIConGPU to run on exascale supercomputers. Largely contributing to refactoring and modernization of code base and documentation.I am also maintaining and… Show more As a postdoc in the computational radiation physics group, I am a core developer and maintainer of C++ exascale particle-in-cell code PIConGPU. My work involves implementing and extending core particle-in-cell routines, including state-of-the-art numerics for incident lasers and field absorbers. I take part in preparing and tuning PIConGPU to run on exascale supercomputers. Largely contributing to refactoring and modernization of code base and documentation.I am also maintaining and contributing to underlying parallel programming libraries alpaka and cupla. Show less

As a software developer in group of prof. Meyerov I further led development of C++ particle-in-cell code PICADOR. My work included generally overseeing code architecture, assisting other developers, co-supervising students.I designed performant parallel primitives for expression of core particle-in-cell operations and extensions. Enhanced concepts and implementation of PIC-MDK interface for portable particle-in-cell extensions, applied it for advanced QED extensions. Led development of… Show more As a software developer in group of prof. Meyerov I further led development of C++ particle-in-cell code PICADOR. My work included generally overseeing code architecture, assisting other developers, co-supervising students.I designed performant parallel primitives for expression of core particle-in-cell operations and extensions. Enhanced concepts and implementation of PIC-MDK interface for portable particle-in-cell extensions, applied it for advanced QED extensions. Led development of a new kernel library for particle-in-cell simulations based on experience gained in PICADOR. Show less

As a junior researcher in group of prof. Meyerov (in parallel to my PhD studies), I continued working on particle-in-cell code PICADOR and led code development. My work included implementing advanced particle-in-cell numerics, including charge-conserving current deposition, incident lasers, and absorbing boundaries. I developed a suite of integration and system tests, and set up nightly builds for PICADOR.I contributed to initial design of PIC-MDK interface for portable particle-in-cell… Show more As a junior researcher in group of prof. Meyerov (in parallel to my PhD studies), I continued working on particle-in-cell code PICADOR and led code development. My work included implementing advanced particle-in-cell numerics, including charge-conserving current deposition, incident lasers, and absorbing boundaries. I developed a suite of integration and system tests, and set up nightly builds for PICADOR.I contributed to initial design of PIC-MDK interface for portable particle-in-cell extensions, and implemented it for PICADOR. Co-developed a novel particle-in-cell implementation for Intel Xeon Phi.Our performance study of realistic particle-in-cell applications on Xeon Phi received the best paper award at Russian Supercomputing Days 2015. Show less

As an assistant in group of prof. Meyerov, I took part in initial development of particle-in-cell code PICADOR. My work largely contributed to the initial code design and implementation of most core particle-in-cell numerics. I led heterogeneous particle-in-cell implementation with OpenCL, that was among the first published fully GPU-capable particle-in-cell codes.

Optimization of broad phase collision detection.