Skip to content
View eminsk's full-sized avatar

Block or report eminsk

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
eminsk/README.md

Hi there, I'm M_N_Nik πŸ‘‹

Software Engineer & Systems Developer with an engineering background spanning back to the early 2000s.

Combining decades of foundational systems programming β€” from low-level architectures (x86/x64 Assembly, Win32 API, C) to modern high-performance Python engineering, asynchronous services, desktop applications, and algorithmic data pipelines.

Telegram Badge Email Badge


🧭 Engineering Philosophy & Background

  • Decades of Craft: Writing code since the early 2000s, built on a deep understanding of memory management, operating system internals, and native Win32/POSIX runtime behavior.
  • Hardware-Aware Performance: Bridging high-level Python productivity with bare-metal speed β€” leveraging custom SIMD SSE2 vectorized math kernels, FASM assembly, and zero-copy C extensions.
  • Architectural Discipline: Strong emphasis on clean code separation, strict typing (mypy), modern tooling (uv, ruff), reproducible builds, and native assembly compilation (FASM x64).

πŸ›  Tech Stack & Core Competencies

  • Languages: Python (3.8–3.16, Free-Threaded No-GIL 3.13t–3.16t, PyPy), x86/x64 Assembly (FASM), C / Win32 API
  • AI Agents & Protocols: Model Context Protocol (MCP) JSON-RPC 2.0 Servers (Claude Code, Antigravity, Cursor, Windsurf), Trajectory JIT Compilation, Vector Search
  • Concurrency & Internals: Asyncio, Free-Threaded Multi-Core (PEP 703), SIMD Vectorization (SSE2/AVX2/FMA/NEON), ctypes & C FFI, Inter-Process Communication (IPC)
  • Desktop & UI Engineering: CustomTkinter, Flet, Tkinter, ttkbootstrap, High-DPI scaling, Native Win32/FASM x64 engineering
  • Data, Math & Trading: OpenPyXL, Pandas, TA-Lib (technical analysis & pattern recognition), NumPy, SQLite
  • Web, Scraping & Automation: Playwright, Requests, cURL-cffi, Pydantic, REST APIs, Session & Auth management
  • Quality & DevTools: uv, ruff, mypy, pytest, Conda-Forge / Rattler-Build, Debian/Ubuntu APT PPA, GitHub Actions CI/CD

πŸš€ Featured Projects

  • agentjit β€” Just-In-Time Compiler for AI Agent Trajectories. Compiles flaky, 30-second multi-step LLM workflows into 0.08ms deterministic Python code with zero token cost. Features speculative guards, automatic bailout runtime, free-threaded (No-GIL PEP 703) support, and Native MCP Server (agentjit-mcp). Featured as the #1 Top Story in The Daily Diff (9/10 Depth & Utility Score). Published on PyPI (v0.1.6) and conda-forge.
  • nanorecall β€” 100% Private, Zero-Cloud, Bare-Metal Desktop Memory & Screen Search Engine in <200KB. Runs on standard CPUs with zero telemetry, automatic privacy shielding for password managers, sub-millisecond semantic search via NanoVector, and Native MCP Server (nanorecall-mcp). Published on PyPI (v0.1.3) and conda-forge.
  • nanovector β€” The SQLite of Vector Search & Episodic Memory for AI Agents in ~120KB. Zero dependencies, instant 0.2ms cold start, pure C99 with AVX2+FMA, ARM NEON, hand-crafted FASM x64 assembly kernels, and Native MCP Server (nanovector-mcp). Published on PyPI (v0.1.6) and conda-forge.
  • nanogemm β€” Minimalist, bare-metal AVX2+FMA & ARM NEON matrix multiplication engine in ~100KB for Python, outperforming NumPy by up to 2.8x on CPU for edge neural network inference, and Native MCP Server (nanogemm-mcp). Featured in The Daily Diff (9/10 Depth Score), trending on GitHub (Deep Learning), and published on PyPI (v0.3.8) and conda-forge.
  • yfinance-ta-patterns β€” Institutional candlestick pattern scanner (all 61 TA-Lib patterns with pure-Python fallback), unbiased next-open backtester, AI Confluence Scorer, Asset Economic Calendar, and Native MCP Server (yfinance-ta-mcp). Supports CPython 3.8–3.16 (including No-GIL 3.13t–3.16t) & PyPy 3.8–3.12. Published on PyPI (v0.3.48) and conda-forge.
  • avito-sdk β€” High-performance headless Avito scraping & data extraction SDK with real-time price drop tracking, full item parameter extraction, Playwright cookie refresh, Telegram/VK notification bots, and Native MCP Server (avito-mcp). Published on PyPI (v0.1.5) and conda-forge.
  • xlsx_vievers β€” High-performance spreadsheet library and desktop processor with a headless 129-function formula engine, hardware-accelerated SIMD SSE2 math, Chart Wizard, AutoFilter, Goal Seek, and Native MCP Server (xlsx-viewer-mcp). Published on PyPI as xlsx-viewer-pro (v1.1.0) and on conda-forge.
  • screenvideo β€” High-performance desktop screen recorder and screenshot suite featuring WASAPI loopback audio, real-time H.264 streaming, Native MCP Server (screenvideo-mcp), and a standalone pure x64 Flat Assembler (FASM) native edition (Release v2.0.0).
  • ppa β€” Official APT Repository (PPA) for Debian/Ubuntu packaging agentjit, avito-sdk, nanogemm, nanorecall, nanovector, xlsx-viewer-pro, and yfinance-ta-patterns.
  • awesome-baremetal-ai β€” Curated benchmark leaderboard and directory of ultra-lightweight, zero-dependency C/C++/Rust/Assembly engines for local LLMs, edge inference, vector search, and AI agents.
  • StackOverflowAPI β€” Modern bilingual desktop client for Stack Overflow built with CustomTkinter, Catppuccin palettes, syntax highlighting, offline bookmarking, and native FASM x64 search client.

🌐 Open Source Contributions


πŸ“¬ Get in Touch

Pinned Loading

  1. xlsx_vievers xlsx_vievers Public

    Headless Excel formula engine (129+ functions), Native MCP Server (xlsx-mcp), SIMD SSE2 math engine, and Office Ribbon desktop spreadsheet app

    Python 2

  2. agentjit agentjit Public

    Just-In-Time Compiler for AI Agent Trajectories, Native MCP Server (agentjit-mcp). Compiles multi-step LLM workflows into 0.08ms Python code.

    Python 3

  3. nanogemm nanogemm Public

    Minimalist, bare-metal SIMD & Assembly GEMM engine for Python. Sub-microsecond CPU matrix multiplication & Native MCP Server (nanogemm-mcp).

    Python 3

  4. nanorecall nanorecall Public

    100% Private, Zero-Cloud, Bare-Metal Desktop Memory & Screen Search Engine, Native MCP Server (nanorecall-mcp).

    Python 1

  5. yfinance-ta-patterns yfinance-ta-patterns Public

    Institutional-grade Candlestick & Chart Pattern Scanner with AI Confluence Scoring, Backtesting, Native MCP Server, No-GIL (3.13t-3.16t) & PyPy (3.8-3.12)

    Python 1 1

  6. nanovector nanovector Public

    The SQLite of Vector Search & Episodic Memory for AI Agents in ~120KB. Pure C99, AVX2+FMA, ARM NEON, FASM x64, Native MCP Server.

    Python 7