Kathmandu, Nepal • UTC+5:45
Available for Projects

Avinash Poudel

@shroudbean

I am a systems & automation engineer building resilient backend pipelines, high-frequency scrapers, and headless browser automation. I focus on clean-room architectures, deterministic behavioral models, and single-binary Go/Python services that operate silently at scale.

Featured Engineering Systems

Macro R

Clean-Room Microsoft Rewards Orchestration & Browser Stealth Engine
Pure Go • CDP

A clean-room rewrite from specification designed to automate Microsoft Rewards workflows autonomously without detection. Eliminates Puppeteer, Selenium, and browser drivers in favor of a direct Chrome DevTools Protocol (CDP) engine with physics-based human typing, scroll mechanics, and seed-derived entropy. Includes a multi-tenant gateway that maps users to dedicated Azure B1s cloud nodes with private live noVNC virtual screens.

ARCHITECTURE Pure Go (Zero cgo / No Node)
STEALTH SCORE 100 / 100 on public detection checks
CLOUD MESH Isolated Azure B1s instances + noVNC
ZERO-CAPTCHA POLICY Strict safety signal halting (no CAPTCHA solving)

NEPSE Live Financial Scraper & Market Engine

High-Frequency Nepal Stock Exchange Extraction Pipeline
Python • Asynchronous • Reverse Engineering

Engineered a resilient market data ingestion pipeline for the Nepal Stock Exchange (NEPSE). Reversed proprietary client handshake protocols, decrypted dynamic token payloads, and built continuous collectors for real-time live trading feeds, floor sheets, order books, and sector indices with zero rate-limit blocks.

THROUGHPUT Real-time tick-by-tick order stream
PAYLOAD HANDLING Dynamic token decryption & session renewal
NORMALIZATION High-precision financial schema output
RELIABILITY Sub-second automated retry on market spikes
Equity Analytics Anti-Bot Evasion FastAPI / WebSockets

Astha Canvas

AI-Powered Retail POS & Intelligent Inventory Engine
Full-Stack • Multimodal AI • Offline-First PWA

A production-grade retail management and POS platform built for art supplies commerce. Features multimodal voice and camera product intake powered by Google Gemini, on-device AI background isolation, offline-first IndexedDB persistence with bidirectional Firebase cloud sync, and thermal barcode billing.

AI INTAKE PIPELINE Multimodal Voice & Vision via Gemini API
ON-DEVICE VISION Client-side Wasm AI background removal
RESILIENT PERSISTENCE Offline-first Dexie.js + Firebase cloud sync
CHECKOUT & AUDIT Barcode scanner, eSewa QR & role PIN auth
React 19 / TypeScript Offline-First (Dexie) Multimodal Gemini AI Zustand & Firebase
How I Build Systems
Clean-Room Engineering
Writing from specification without copying legacy architecture, fragile selectors, or bloated npm dependencies.
Deterministic Entropy
Eliminating shared constants across bots and scrapers so actions derive mathematically from unique seeds, impossible to cluster.
Headless Evasion
Driving real Chromium displays over Xvfb rather than vulnerable --headless flags that get flagged by anti-bot classifiers.
Single Binary Simplicity
Compiling to single, statically linked executables with embedded assets and zero runtime dependencies.
$ curl -sL https://avinashpoudel.com.np/status
{"status": "operational", "location": "Kathmandu", "focus": "Systems / Automation"}