Applied Mathematics & Computer Science student at IKBFU specializing in Artificial Intelligence and Data Analysis. I engineer local & hybrid ML pipelines, build responsive desktop systems with reactive UI/UX, and develop mathematically grounded simulators for academic and applied domains.

• 🌟 Akira — Autonomous multimodal desktop assistant & AI companion.

  • Hybrid LLM Architecture & Semantic RAG: Local vector retrieval via FastEmbed (ONNX) over a multi-chapter character bible, combined with intelligent dynamic model routing (OpenRouter / DeepSeek / Claude / GigaChat) and long-term async memory extraction.
  • Voice & Vision Subsystems: Low-latency local speech pipeline (STT/TTS via INT8 ONNX models with GPU DirectML acceleration), real-time face verification (Face ID), fatigue tracking, and gesture recognition via MediaPipe.
  • Reactive Desktop Workspace: Built on PyQt6 featuring three adaptive interaction modes (ambient Mini-widget, floating Companion, and wide Studio Workspace with anchored animation curves). Includes single-instance IPC sockets and Telegram remote control.
  • Packaged into enterprise-grade installers using WiX Toolset (MSI).

• 🎛️ Reis105M1 Trainer — Interactive educational simulator for digital pulse reflectometers.

  • Engineered with C# / .NET 8 (WPF) for university labs to train students in cable reflectometry and line fault diagnostics (open/short circuits).
  • Custom 128×64 monochrome LCD display engine rendering real-time vector reflectograms, coordinate interpolation grids, and live parameter calculations.
  • Vector-driven animation physics for line commutation, BNC connectors, and measurement probe clamping.

• 🃏 Velaris — Modular Python card game ecosystem & multiplayer network.

  • Custom low-latency socket networking handling concurrent multiplayer sessions (Blackjack, Durak, Texas Hold'em).
  • Event-driven GUI architecture with standalone multi-platform distribution binaries.

• 🗺️ Military Topography Training Suite — Educational software package for geodesy and automated knowledge grading (datacwtopograf, datalyali, datatopograf).

  • Direct and inverse geodetic calculations: directional angles, magnetic declination/azimuths, and coordinate conversions (Geographic $\leftrightarrow$ Gauss-Krüger).
  • Built using Delphi / Object Pascal with automated student testing and error-margin validation.

• 👁️ Computer Vision & Kinematics Tools — Real-time posture tracking, ergonomic landmark analysis, and spatial geometry using OpenCV, MediaPipe, YOLO, and PyTorch.

Languages:

AI, Machine Learning & RAG:

Desktop Engineering & Packaging:

• Hybrid AI & Routing: Designing local-first RAG pipelines with semantic vector indexing, zero-latency caching, and intelligent API model fallbacks (lowering token consumption up to 10×).
• Embedded & Desktop Optimization: Running speech-to-text, neural TTS, and computer vision models on consumer hardware via DirectML/INT8 quantization.
• Robust GUI Architecture: Building multi-threaded desktop software with non-blocking async workers, single-instance socket guards, and fluid kinetic animations.

"Turning complex algorithms into clean architecture and reliable tools." 🐾