Ivan Dryga

- I built a low-latency market data collection service (including tick-level data) from exchanges such as Bybit and Binance in Go;
- I developed a processing service for the incoming exchange data (resampling, merging multiple time series, computing indicators, and fusing inputs from multiple models) in Rust;
- I designed and deployed ensemble neural services for working with time-series and news data, as well as agent-based services that place SL and TP according to defined strategie;

I continue to optimize time-series agents and models at a low level for 4× RTX 5090 and H200, respectively. LLM workloads require more memory, so optimization is primarily done on H200, while agents must make decisions quickly, so they are tuned on the RTX 5090 due to its stronger compute throughput. The goal is to achieve HFT-grade latencies using graph optimization, mixed precision, ONNX Runtime, TensorRT, inline PTX, Triton, and kernel fusion. For profiling, I use torch.profiler, nvidia-smi, Nsight Systems, and Nsight Compute.

Development of Strategic and Tactical Maps for the Ministry of Defense, and Civil-Use Maps with API Integration (Flightradar, MarineTraffic, Earthquake)
(C++20, Vulkan API, RHI, CMake, Android, Windows, Linux)

- Designed and maintain the core architecture of the rendering engine and application back-end;
- Implemented a dynamic resource streaming system for rendering needs based on an LRU cache;
- Built a custom render engine on top of Vulkan API;
- Created an adaptive multi-threaded pipeline for loading, rendering, and preparing resources and data required for drawing;
- Developed an asynchronous system for loading, rendering, and resource preparation using double-buffering, condition variables, spin-locks, and lock-free structures (all stages run in parallel, and each step generally has the duration of one frame);
- Implemented multi-threaded rendering with Vulkan API;
- Built a GeoJSON/JSON parsing and processing module for use inside the render engine;
- Integrated ingestion, processing, and visualization of live data from platforms such as Flightradar, MarineTraffic, and Earthquake feeds;
- Added wind and weather map layers;
- Developed a terrain-rendering system based on height- and depth-maps via a Vulkan compute pipeline;
- Implemented “lazy” on-the-fly streaming and rendering of resources;

- Adapted the A* algorithm to support maritime and road routing.Built a streaming-data collection subsystem that ingests external platform data and stores it in a database;

- Created an asynchronous frame-preparation pipeline that packages data for rendering. (Kubernetes, Docker Compose, Kafka, Go, Grafana, PostgreSQL).

I developed a real-time video processing service that uses Vulkan shaders together with FFmpeg and GStreamer (decoding individual frames and processing them with Vulkan shader pipelines).

- Developed an adaptive and universal “plasticine” algorithm (filling gaps between teeth) (patented by the company);
- Developed a universal algorithm for eliminating gum recession (patented by the company);
- Developed an ML-based 3D human skull modeling system using CT scans in DICOM format;
- Developed an internal desktop model editor to streamline the workflow for applied specialists
- Developed a simple rendering engine using OpenGL and C++20.

- Development the new video/photo-editor engine based C++ and OpenGL;
- Profiling and optimization of the entire rendering pipeline in the engine's internal level with Nvidia Nsight System and Android GPU Inspector (AGI);
- Development of an application for automated testing the new video/photo-editor engine based C++;
- Support old engine that based on Java/Kotlin and OpenGL;
- Contributed to the development of architecture video/photo-editor engine;
- Development application for automation testing the new video/photo-editor engine based C++;
- Built an ingest service in Go to receive RTSP streams and route stream sessions to the orchestrator for downstream distribution;
- mplemented an orchestrator service that can switch FFmpeg and GStreamer workers on the fly;
- xtended FFmpeg- and GStreamer-based workers to support multi-variant transcoding, enabling native (non-AWS) delivery of streams at multiple resolutions and bitrates;
- Developed low-latency video delivery modules based on LiveKit to stream over the WebRTC protocol;
- Built a viewer delivery module based on LL-HLS.

- Work mainly with theoretical application layer graphics, with in order to show their DirectX-oriented rendering routines in a more Vulkan-friendly shape;
- Extending the public proprietary API to include new features (such as, for example, mesh, shaders, ray tracing, volumetric lights, deffered rendering etc.);
- Profiling and optimization of the entire rendering pipeline both in the application, as well as at the internal level;
- Engine support for natural reality devices;
- Participation in creating the new philosophy of the graphic's engine focused on GPU-driven rendering.

- Developed physics engine for interactive education system, enhancing employee training experiences;

- Contributed to architecture design of both game engine and rendering engine;

- Collaborated with outsourced professionals to define and achieve animation objectives;

- Managed team of 3 programmers, ensuring project milestones were met;

- Engineered graphics solution utilizing OpenGL API, implementing IBL PBR, SSAO, ray tracing, and volumetric lighting technologies;

- Applied narrative directions in creation of motion graphics effectively;

- Constructed core backend system using PHP and MySQL with MVC architecture for infrastructure control;

- Demonstrated strong work ethic and commitment to professional development.