Independent Technology Lab · Open-Source Tools

Useful software.
Open by default.

Hukam Labs is an independent open-source lab building practical AI, creative, developer, and automation tools.

We engineer standalone desktop applications, browser extensions, and local AI pipelines to solve actual production bottlenecks — freely available without forced subscriptions, cloud telemetry, or artificial paywalls.

Explore Projects GitHub Profile /llms.txt ↗

16 Open-Source Tools

Local-First On-Device Inference

0 KB Telemetry or Paywalls

Permissive MIT & GPL Licensed

A collection of things I've built.

Every utility in this lab began as an answer to a real production constraint — and was rebuilt into clean, documented open-source software for the community.

Flagship Extension Active · v6.3.1 Chromium MV3

Hukam Flow

Production-ready Chrome Manifest V3 workflow automation tool for Google Flow (flow.google.com). Features a dedicated dockable Side Panel, MAIN-world page hooks for Angular trusted events, real-time API response interception, and prompt-indexed auto-downloading.

The Real Problem Solved

Automating prompt batches on Google Flow fails because Angular cancels synthetic DOM events, network timeouts break long runs, and manual downloads lose prompt associations.

Dedicated Chrome Side Panel UI (Alt + Shift + F)
MAIN-world Angular event synthesis hook bypassing form validation barriers
Real-time interception of internal Google Flow batchexecute network APIs
Smart Auto Mode: Auto-run, wait, auto-download, with prompt-indexed file naming

Chrome MV3 Side Panel API Angular Hooks Web Request Zero Telemetry

Project Details → Repository Releases ↗

Side Panel Queue Runner Tab #241 Active

MAIN-world Angular Hook WAITING

Synthetic Prompt Dispatch WAITING

batchexecute Interception WAITING

Prompt-Indexed AutoDownload WAITING

// Idle. Click "Run Batch Demo" to simulate Chrome Side Panel queue.

Standalone Executable Active · v1.0.0 Vulkan NCNN

Hukam AI Upscaler Pro

Hardware-accelerated Vulkan Real-ESRGAN super-resolution desktop app for Windows x64. Bundles pre-converted weights into an ultra-portable 55MB single-file executable with resume logging and tile-based memory safety.

The Real Problem Solved

Super-resolution deep learning tools usually require 4GB+ of PyTorch/CUDA libraries (locking out AMD/Intel users) or expensive cloud subscriptions that leak sensitive creative artwork.

Hardware-accelerated Vulkan Real-ESRGAN running across NVIDIA, AMD, and Intel GPUs
Standalone 55MB executable with pre-bundled neural weights
Crash-resilient resume log surviving power cuts and batch pauses
Low-VRAM tile splitting preventing Out-Of-Memory crashes on laptop GPUs

Real-ESRGAN Vulkan NCNN Python C++ Windows x64 Zero Cloud

Project Details → Download .EXE (~55MB) GitHub ↗

Original (Uncompressed Crop) Vulkan 4x Enhanced Crop

Vulkan 4x Real-ESRGAN

In-Browser AI Active · v1.0.0 WebGPU Accelerated

SVG Magic Cleaner

100% private, browser-based AI background remover and raster-to-SVG vectorizer using Hugging Face Transformers.js and WebGPU. Executes briaai/RMBG-1.4 locally with zero cloud uploads.

The Real Problem Solved

Graphic designers frequently need transparent cutouts and vectorized logos, but existing tools upload client files to remote servers or bill recurring per-image credits.

In-browser local inference via WebGPU using briaai/RMBG-1.4
Automated raster contour tracing into pure SVG Bezier paths
100% client-side privacy: zero images sent over the network

React 18 TypeScript WebGPU Transformers.js SVG Tracer

Project Details → Repository

Raster Mode · Transparent PNG

Execution: Local WebGPU 42 Bezier Curves 0 KB Cloud

Binary Forensics Active · v1.1.0 Lossless Reconstruction

AI Image Forensic Repair

Lossless pixel buffer reconstruction utility solving the infamous Adobe Illustrator "File format cannot be placed" error on AI images (Midjourney, ComfyUI, DALL-E) by rebuilding corrupt container headers.

The Real Problem Solved

Generative AI engines frequently generate non-standard PNGs with truncated chunks or unsynchronized markers that cause professional Adobe vector software to fail to import.

Synthesizes valid ISO standard PNG signatures and mandatory IHDR chunks
100% lossless: preserves original RGB/RGBA raster buffer without recompression
Batch directory processing with automatic unplaced file recovery

Python Pillow Binary Struct PNG Specifications Standalone .EXE

Project Details → Repository

Binary Header Inspector PNG Container

Error: Adobe Illustrator throws "File format cannot be placed"

00 00 00 00 00 00 00 00 · NULL Padding / Missing Magic Signature 49 48 44 52 00 00 00 00 · Truncated IHDR chunk descriptor ?? ?? ?? ?? ?? ?? ?? ?? · Unsynchronized pixel buffer markers 00 00 00 00 00 00 00 00 · Missing pHYs / Corrupted color profile

ISO/IEC 15948

Explore The Lab

The Complete Open-Source Index

Browse all 16 production tools, AI models, and creative utilities engineered across Hukam Labs.

Inside the Lab

Independent Software Craft

Hukam Labs is an independent technology workshop where Waleed Ahmad ("Hukam") experiments with local AI inference, desktop toolchains, and browser automation to build practical tools that solve real bottlenecks.

Rather than building speculative commercial SaaS applications with forced subscriptions, the lab focuses on standalone, executable software that creators and engineers can run locally on their own machines without external surveillance or cloud friction.

Origin in Daily Bottlenecks

Every utility here was born out of an actual production frustration: removing backgrounds from hundreds of graphics without recurring API fees, running super-resolution without NVIDIA CUDA driver conflicts, or repairing malformed image files for Adobe Illustrator.

When a tool proves effective in production, it is cleaned, documented, licensed under MIT or GPL, and published openly for the developer and creator community.

01 // CONCEPT

Production Constraint

Identified during daily design, automation, or creative development workflows.

02 // PROTOTYPE

Local Experimentation

Rapid iteration testing local Vulkan kernels, browser DOM hooks, or WebGPU shaders.

03 // ACTIVE

Production Hardening

Built into self-contained desktop apps, Chrome MV3 extensions, or CLI pipelines.

04 // COMPLETE

Open Publication

Published openly under MIT or GPL with full documentation and zero telemetry.

Built to Be Used

Core Engineering Principles

Useful software should be accessible. We believe practical engineering speaks through executable code, transparent repositories, and measurable workflow speedups.

01 // LOCAL FIRST

Tools execute directly on your CPU, GPU, or browser runtime. Prompts and media never touch remote servers.

02 // ACCESSIBLE

No subscription paywalls, artificial compute tokens, credit expiration, or registration gates.

03 // TRANSPARENT

Code is published under permissive MIT or GPL licenses. Anyone can inspect, audit, or fork freely.

04 // INDEPENDENT

Maintained independently without venture capital pressures or artificial marketing hype distorting UX.

Repository Inspector Direct GitHub Mirror

Engineering Lead

Waleed Ahmad

"Hukam" · Independent Developer & AI Specialist · PSEB Registered Freelancer

Waleed Ahmad is an independent software developer, AI pipeline specialist, and PSEB registered freelancer based in Pakistan. He architects desktop utilities, Vulkan GPU pipelines, and browser automation systems.

His personal commercial portfolio (likehukam.com) showcases client engagements and technology advisory work, while Hukam Labs represents his independent open-source research and freely accessible software contributions.

GitHub Profile likehukam.com ↗ [email protected]

LAB PROFILE

Handle: @hellohukam

Status: PSEB Registered

Studio: Hukam Labs

Inquiries: Email directly