Skip to content
View eminsk's full-sized avatar
6 followers · 0 following

Achievements

Achievement: Pull Sharkx2Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pull Sharkx2Achievement: QuickdrawAchievement: YOLO

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. 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 👋

Systems Engineer & Compiler Developer with an engineering background spanning low-level architectures (x86/x64 Assembly, Win32 API, C) to modern high-performance Python runtime engineering, AOT compilation, and native SIMD acceleration.

Developer Portal Conda Downloads Telegram Email

🧭 Engineering Philosophy & Background

  • Decades of Systems Craft: Writing software since the early 2000s, grounded in memory management, CPU microarchitecture, hardware registers, and native Win32/POSIX runtime behavior.
  • Hardware-Aware Performance: Bridging high-level productivity with bare-metal speed — developing native AOT compilers, custom SIMD (SSE2/AVX2/NEON) vector kernels, FASM x64 assembly engines, and zero-copy C extensions.
  • Architectural Discipline: Strict typing (mypy), modern tooling (uv, ruff), reproducible packaging, hardware concurrency, free-threaded Python (No-GIL PEP 703), and native ABI calling conventions.

🛠 Tech Stack & Core Competencies

  • Compilers & Low-Level Systems: Python AST to pure x86_64 FASM AOT Compiler, Peephole Optimization, Calling Conventions (Microsoft x64 ABI / System V ABI, 16-byte stack alignment, shadow space), CPU register allocation, hardware atomic primitives (lock add/sub).
  • Languages: Python (3.8–3.16, Free-Threaded No-GIL 3.13t–3.16t, PyPy), x86/x64 Assembly (Flat Assembler - FASM / FASMG), C99 / Win32 API
  • High-Performance Computing (HPC) & SIMD: SIMD Vectorization (SSE2/AVX2/FMA/ARM NEON), BLAS / GEMM matrix multiplication micro-kernels, SIMD JSON parsing, Vector Search & Embeddings
  • Concurrency & Multithreading: True No-GIL Native OS Threads (CreateThread), Free-Threaded Multi-Core (PEP 703), Lock-free & Atomic synchronization, Asyncio, Inter-Process Communication (IPC)
  • AI Systems & Runtime Protocols: Model Context Protocol (MCP) JSON-RPC 2.0 Servers, Agent Trajectory JIT Compilation, Semantic Vector Search
  • Packaging & Infrastructure: uv, ruff, mypy, pytest, Conda-Forge / Rattler-Build, Debian/Ubuntu APT PPA, GitHub Actions CI/CD pipelines

🚀 Featured Projects

  • pureapk (pyapk) — Zero-dependency pure Python Android APK engine with Android 15 16KB ELF page alignment (tag 0xd935), APK v1/v2/v3 signatures, multi-store packaging (APK, Google Play AAB, iOS IPA stub, Windows 11 MSIX), in-place APK modification, and Android wheel extraction without Java or Android Studio. Published on PyPI (pureapk v1.0.2) and available on Debian/Ubuntu APT PPA (python3-pureapk).
  • pyfasm — Universal Native No-GIL (3.13t–3.16t) & PyPy to pure x86_64 Flat Assembler (FASM) Native Compiler. Translates Python AST directly into self-contained standalone PE+ (.exe, .dll), ELF64, and Mach-O binaries (~9 KB) with zero virtual machine and zero runtime overhead. Features true multi-threaded native concurrency (CreateThread with hardware atomics), SSE scalar/vector floating point, SIMD-accelerated JSON parser, and native Win32 GUI engine. Published on PyPI (pyfasm64) and announced on the Flat Assembler Board.
  • fasm-modern-sdk — Modernized SDK and header toolchain for Flat Assembler (FASM v1.73 / FASMG) featuring 64-bit Windows equates, official calling conventions, and zero-configuration bindings.
  • agentjit — Just-In-Time Compiler for AI Agent Trajectories. Compiles 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 and conda-forge.
  • nanovector — 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). Over 7,900+ downloads on Conda. Published on PyPI and conda-forge.
  • nanogemm — 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). Over 9,800+ downloads on Conda. Featured in The Daily Diff (9/10 Depth Score), trending on GitHub (Deep Learning), and published on PyPI and conda-forge.
  • nanorecall — 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 and conda-forge.
  • yfinance-ta-patterns — 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. Over 1,900+ downloads on Conda. Published on PyPI and conda-forge.
  • xlsx_vievers — 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 and on conda-forge.
  • screenvideo — 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.
  • ppa — Official APT Repository (PPA) for Debian/Ubuntu packaging pureapk, agentjit, nanogemm, nanorecall, nanovector, xlsx-viewer-pro, and yfinance-ta-patterns. Available at eminsk.github.io/ppa.

🌐 Open Source & Upstream Engineering

📬 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

    Assembly 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 4 1

  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. 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

  5. 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

  6. avito-sdk avito-sdk Public

    High-performance headless Avito scraping & data extraction SDK with price drop tracking, Playwright cookies, Telegram/VK bots, and Native MCP Server

    Python 4 1