Software Engineer · Iran — Remote / Worldwide

Mohsen Taghipour

Software Engineer

I build systems from scratch — from Wi-Fi packet parsers on raw sockets and LLMs on 4 GB VRAM to Wayland compositors in pure Rust. No wrappers, no black boxes.

13Projects & apps
6Languages & tools
24+years old
MTneganinshell
Software Engineer · B.Sc. Software Engineering
“I open black boxes.”
~85k
LOC (MTP)
<1.5s
boot → desktop
127 MB
VRAM peak
viyad.ir
live SaaS
About

Engineer who opens black boxes

24, B.Sc. Software Engineering. I ship low-level, high-performance systems — the kind where you read the spec, open the black box, and write the replacement. 13 projects live here: a Rust LLM trained on a laptop GPU (4 GB VRAM, 127 MB peak), a full Wayland desktop suite that boots in <1.5s at ~185 MB idle, a live classroom SaaS (viyad.ir) with WebRTC + LiveKit, and a Wi-Fi recon suite that parses 802.11 by hand.

Stack I actually use: Rust (CUDA/cuBLAS, Wayland/Smithay, Tokio), Python (Flask/aiohttp), Kotlin (Compose, Clean Architecture), TypeScript/Svelte 5 + React 19, Bash/Linux, SQLite/Redis/Mongo. Every project page documents architecture, file layout, dependency map and exact run instructions — not marketing copy.

Curiosity is the skill. If it looks like magic — radio frames, tensor cores, Wayland protocols — I take it apart until it isn’t. Then I ship it.

Read more about me

Skills & Toolbox

Every project uses a different slice of the stack — from CUDA kernels to Compose — one philosophy: from scratch, clean, explainable.

Languages

RustPythonKotlinTypeScript / JavaScriptBashSQL

Systems & Low-level

CUDA / cuBLASWaylandLinuxSockets / Raw PacketsConcurrency

Frameworks & UI

Svelte 5Jetpack ComposeTauriRatatuiFlask

Tools & Platforms

GitAndroidWebSocketPWASQLiteHyprland
Featured

Projects

All projects →

★ Featured

★ FEATURED

LucilleAI

★ FEATURED
🧠AI / Deep Learning

A full deep-learning framework in pure Rust: autograd, SIMD, cuBLAS/Tensor-Core kernels and a Persian Unigram tokenizer, used to pre-train and SFT an LLM on a GTX 1650 Ti.

RustCUDA / cuBLASTensor Cores (FP16/AMP)AVX2 SIMDTokio+2
View details
★ FEATURED

MTP Ecosystem

★ FEATURED
🖥️Desktop & Linux

Five cooperating Rust projects (~324 files, ~85k lines, zero C dependencies): a GPU-accelerated Wayland desktop that boots to a usable session in under 1.5 seconds at ~185 MB idle RAM.

RustWaylandWGPUDRM/KMSSmithay+4
View details
★ FEATURED

Rsut (Rust Monitor)

★ FEATURED
🛡️Security & Networks

Rust + Svelte 5 Wi-Fi monitoring/recon tool: raw 802.11 sniffer, PMKID/SAE/4-way-handshake capture engines, channel hopping, live WebSocket dashboard with real-only data and an attack/toolbox panel for security research on your own network.

RustSvelte 5SvelteKitWebSocketlibc (raw sockets)+2
View details
★ FEATURED

AURA

★ FEATURED
🖥️Desktop & Linux

A complete theming engine + desktop environment configuration for Arch/Hyprland: the Aura Engine regenerates every component (Hyprland, Waybar, Rofi, Kitty, SwayNC, Hyprlock, Fastfetch) from a single theme.json, with dynamic wallust-based theming and 23 curated themes.

BashHyprland (Wayland)WaybarRofi 2.0Kitty+2
View details
★ FEATURED

Negan Agent

★ FEATURED
🧠AI / Deep Learning

A terminal-first AI coding agent written in Rust (~2,800 lines): a cognitive core with planning, reflection, a code-quality judge, self-critique, persistent skill memory, a replay engine and a safety sandbox — served through a Ratatui TUI, a WebSocket web UI and local models via Ollama.

RustTokioAxum + WebSocketRatatui / crosstermollama-rs+3
View details

All projects

MagicWatch

📱Mobile2025

A Kotlin + Jetpack Compose Android app built with Clean Architecture: prayer-time engine (Tehran/ISNA/MWL methods), azan + Qur'an audio via Media3, Jalali calendar, always-on display, floating clock, tasks, weather and 4 themes — with Room, DataStore, Hilt and Retrofit.

KotlinJetpack ComposeClean ArchitectureHiltRoom+3
View details

IranTop

🛠️Tools & CLI2025

A real-time Linux system monitor in Rust + Ratatui: 4 chart modes (line/bar/sparkline/donut), split-screen CPU view, live process manager with search & kill, NVIDIA GPU monitoring and full mouse support — lightweight by design.

RustRatatuiCrosstermsysinfoLinux
View details

Transfer

🛠️Tools & CLI2025–26

A Python LAN file-sharing server (aiohttp + WebSocket + HTTPS) with drag & drop, ZIP multi-download, QR pairing, device discovery, 5 themes, bilingual UI, PWA support, remote camera/screen viewing, chat, speed tests and rate limiting — packaged into a signed Windows installer and a CLI tool.

Python 3.10+aiohttpWebSocketPWAPyInstaller+2
View details

Beacon Chat

🌐Web Apps2025

A private LAN chat system (Python: Flask + WebSocket + SQLite) with no internet and no cloud: DMs, group rooms, file/media uploads up to 60 MB, emoji avatars, typing indicators, edit/delete, online list, browser notifications and full persistence across restarts.

PythonFlaskWebSocketSQLiteHTML/CSS/JS+1
View details

Grub — Matrix/Cyberpunk Theme

🖥️Desktop & Linux2025

A production-ready GRUB2 gfxmenu theme: 9-patch frosted-glass menu, glowing cyan selection, terminal typography, an ImageMagick binary-rain compositor and an idempotent installer with safe GRUB defaults injection and clean uninstall.

BashGRUB2ImageMagickLinux
View details

Sajjad Inventory

🌐Web Apps2025

A Persian RTL web app with local storage (IndexedDB) and a service worker: user login, item & category management, stock-in and dispatch workflows, full history, low-stock alerts, Chart.js inventory charts, Excel export and complete offline support.

HTML/CSS/JSIndexedDBPWA / Service WorkerChart.jsSheetJS+1
View details

Fajr Danesh School Platform

🌐Web Apps2025

A production school platform for Fajr Danesh: student/teacher/parent/admin dashboards, news, gallery, progress charts, quizzes, class schedules, smart notifications and role-based access — React 18 + Vite + Capacitor on the front, Express + Mongoose on the back, running live at fajrdanesh1.ir.

React 18ViteCapacitor (Android)Node.js / ExpressMongoDB / Mongoose+5
View details

Viyad

🌐Web Apps2026

A production online-classroom platform: real-time WebRTC video with STUN/TURN, an interactive whiteboard, realtime PDF materials with auto font-fixing, quizzes, chat, screen share and recording — a modular Node.js/Express backend (500+ files) and a React 19 FSD frontend, live at viyad.ir.

Node.js 20 / Express 5React 19 / Vite 7MongoDB / MongooseRedis + Socket.io (Redis adapter)WebRTC / LiveKit / STUN-TURN+6
View details