Why would one need to parse BEAM (Erlang virtual machine files) in Python language? The reason is called rpmlint. This tool is employed by a vast majority of modern RPM-based Linux distros to ensure RPM packaging quality. When packaging application, written in Erlang, an ability to extract metadata from just compiled *.beam files may come in handy. As an example, pybeam is used to check debug_info presence at openSUSE, or to provide automatic RPM package dependency generation for Erlang at Red Hat.
Under the hood, pybeam heavily relies on construct — declarative binary parser.
OpenVizsla is an open hardware project aimed to provide USB debugging and sniffing tool. It consists of USB PHY connected to FPGA, equipped with FTDI chip to communicate with a host PC. Even though the Python-based host software is provided by the hardware project authors, libopenvizsla provides an alternative plain C interface to communicate with the device. Later, Tomasz Moń implemented and contributed WireShark external cap based on the library. The external cap makes it possible to capture USB frames and analyze them with WireShark dissectors. There is also a module for libpcap, the famous packet capturing library used by tcpdump and WireShark.
FITS is a legacy format for storing raw data obtained on astronomical telescopes. Even yet the format is forty years old, it is in heavy use in a modern astronomy world. Internally, it is like a numpy array dump format. So, Fips project is about copying a FITS file as-is into the GPU memory and employing different dirty OpenGL hacks to show this memory as an image. Surprisingly, this approach has drawn the attention of astronomical software community. Fips is a joined effort with my friend and colleague Konstantin Malanchev.
Coniferest is a Python package for active anomaly detection, developed collaboratively by Konstantin Malanchev, Vladimir Korolev, and the SNAD team. Built for performance, it offers a Rust-powered reimplementation of the classic Isolation Forest algorithm for faster scoring, along with an enhanced version of the Active Anomaly Detection (AAD) algorithm and the team's own innovative PineForest model for active learning.
An open-hardware design project for a USB microSD card reader, built on the widely available Realtek RTS5170 controller. Designed from the ground up in KiCad for educational, prototyping, and manufacturing purposes, its primary technical utility is achieved on Linux systems, where the device works with standard kernel drivers, presents the card as a standard /dev/mmcblk* block device, and—crucially for developers and hardware enthusiasts—allows the direct exchange of raw MMC commands with the card through this interface, enabling low-level testing and communication.