Dmitry Ursegov Email & Phone Number

Moscow, Ru

Moscow, Moscow City, Russia

Building distributed PostgreSQL.

Moscow, Moscow City, Russia

Reorganized and build the team that worked on core and systems components for all products of the company. The team was mainly responsible for the Greenplum components:• Built MVP of Greenplum Web UI, demonstrated and installed it for one of the biggest Russian food retailer. The first version included animation of the plan tree execution in real-time, allowed to kill queries and check the history. I was in charge of architecture planning, technical backlog and review process.• Developed Greenplum to Clickhouse connector. My contribution was in performance optimizations that resulted in a top speed per segment by eliminating extra data copy and conversion.• Worked with support team on L3 tickets and pushed fixes upstream.

The start-up created the world’s first GPU-native software engine specifically designed to fully leverage the massively parallel processing power of GPUs for stream and hyperbatch processing. It is currently focusing on further improvements to the engine and its integration into the popular Apache Spark framework.• Developed a prototype of the technology for the seed funding.• Built a team of 6 people, having conducted a few dozens of interviews. Later the team increased to 10 people.• Played a key role in selecting the technological stack and designing the architecture, reducing the time of MVP development thanks to the open-source products and dividing the development into two independent components: modified Apache Spark and a GPU-native engine written in C++.• Was in charge of the technical leadership of the engine development and coordinated the work on Apache Spark.• Acted as a scrum-master at the initial stage. Formulated the definition of done rules, which boosted the code quality thanks to the review, better test coverage and testing automation. Used agile metrics to better estimate the release time.• Was in charge of performance: performance modelling, debugging and profiling. Having used popular Yahoo and Taxi Rides tests, we obtained 5-10 times better results than on similar-priced CPUs. For computationally intensive problems, the difference was tens of times.• Implemented computational model of Apache Spark on GPU, made the engine thread-safe. The tasks were performed in separate CUDA Streams. Added Multi-gpu support.• Implemented a local shuffle for a single node configuration which was used as a part for multi-node implementation.• Implemented the conversion of CSV into columnar format on GPU.Technologies: Agile, Apache Spark, MapReduce, C++, CUDA, SQL, LLVM, NUMA, Linux kernel, perf, async-profiler, nvvp, asan/tsan, git, gdb

CounterTack project, USA, countertack.comThe company develops solutions in cybersecurity field. It was one of the first which suggested collecting system events from workstations and analyzing the behavior for identifying attacks on the infrastructure level (Endpoint Detection and Response). Platforms: Windows, Linux, MacOS.• Participated in development of an agent for Windows workstations. The agent works in the kernel space and is engaged in interception of system events, their preliminary analysis and sending aggregated information to the server.• Optimized the performance of event handlers and reduced their work time by several times.• Ported the agent to Linux and MacOS platforms and supported them.• Made basic code independent of the platform making it easier to support and add new features.• Mentored new developers and reviewed the code.Technologies: kernel level and system programming, C, Windows Performance Toolkit, gperftools, valgrind.

WindRiver project, USA, windriver.comThe development of a dynamic tracing tool (sensorpoint). Linux and VxWorks platforms. X86, ARM, PPC, MIPS architectures.• Optimized the performance and memory consumption.• Worked on ELF files parsing and dynamically modification of structures, data and code in memory.• Fixed difficult bugs related to parallel execution, memory model, built-in assembler, etc.Technologies: kernel level and system programming, C, gperftools.

Moscow, Russian Federation

A project to use Xen ARM hypervisor to run applications from other operating systems on Android.• Worked on the framework modification for integrating third-party applications.• Contributed to the development of a prototype of integration with the Qtopia platform.Technologies: Android framework, Java, C++

Moscow, Russian Federation

Development and support of the Lynx operating system.• Developed a network card driver.• Fixed bugs in drivers and the kernel.• Provided technical support and documentation.Technologies: kernel level and system programming, C.

Moscow, Russian Federation

Client-side work, administration of SUN servers and Solaris operating system.Technologies: system administration, Shell scripting, TCP/IP networks.

Student, Mathematics

Specialist, Radiotechnicks