I was tasked to help support a legacy product. Debugging and fixing existing bugs, adding support for new features/hardware revisions, refactoring/generally improving existing code and adding documentation.
The product is a wearable device plus its docking station (which supports many devices at the same time). Both devices run a custom Yocto-built embedded Linux distro, where the dock has a Qt GUI.

The main project challenges were:

• Limited to no documentation
• Documentation that did exist was in Spanish and/or out of date
• Unsafe and extremely unclear/convoluted code with tons of magic values
• Adding new features whilst supporting multiple hardware revisions

Some of my achievements in this role:

• Adding documentation/static analysis/formatting rules/tests, whilst minimizing logic changes (reducing regression risk)
• Setting up actual support for debugging (valgrind + gdb(server))
• Updating build process to use GNU Make over fault-sensitive magic lines of 'just copy from the readme', updating the CI/CD to use the same tooling
• Add feature to update tool to reduce >20 min run time to <1 min
• Add support for WPA3 to the existing old kernel by backporting the functionality
• Adding various debug tools: VNC/helper scripts/quality of life improvements

I was tasked to help develop a new type of 'person position/state sensor' for the client (who wishes to remain anonymous) based on thermal imaging and a nRF52 microcontroller.
Primarily focusing on the improving the existing software and making it so it can be re-used for different projects. The main project challenges were:

• Working with limited documentation (+ not having access to the engineer(s) who wrote the code)
• Unclear and unsafe code (causing frequent hard to debug failures once they were not just created/tested in isolation)
• Working with not fully validated HW+SW -> debugging often required logic analyzers and close communication with hardware engineers
• I also had input on hardware changes to minimize the power consumption
• I was the only C++ engineer -> was working very independent
• Including:
• Setting up doxygen for documentation
• Improving the build system
• Being more structured/clear
• Being reusable (with git sub-modules)
• Creating a GUI in Qt for testing
• Creating a lot more structure/order within the software.
• Project specifications were not finalized -> I was involved in a lot of discussions about what was actually desired/required and how to achieve that.

Arma 3 is a multi-player military shooter “simulator” (similar-ish to Call of Duty). It allows servers to install mods (short for modifications. Extending the existing game with new maps/models/game types etc).

I setup a server so my friends and I could play together. The problem was that they wanted to play a lot more than I did and kept asking me to change the mods on the server/restart it. Asking them to learn some basic Linux and SSH into my server and fix it themselves wasn't feasible. So I made a web front-end for them to use.

This project effectively had 3 parts:

• The actual arma server; which I already scripted together in Bash
• The web front-end (written in Angular 5)
• The web back-end (written in C++)

The front-end effectively has no decision making, it simply relays button presses/file changes to- and receives updates from the back-end. User authentication is handled by Authelia, which gives the user's groups to the backend; which then authorizes actions. Communication between the front-end and back-end uses WebSockets. A functional (excusing the dev-art) front-end was made and used.

Plans were made to update the UI to be a lot more user-friendly and prettier to look at, but the friend group stopped playing the game before that happened -> it was never fully updated. It was to be built using the Tailwind CSS framework.