Question 1

Do you take over existing codebases, or only start new projects?

Accepted Answer

Both. A lot of desktop work is on applications that already have users. I start with a short assessment — building the app on a clean machine, reading the architecture, running it under a profiler and a crash reporter — and then propose changes in an order that keeps the app shippable the whole time.

Question 2

Should our new Windows app use WPF, WinUI or something cross-platform?

Accepted Answer

It depends on where it must run and who will maintain it. For Windows-only, data-heavy tools, WPF on modern .NET is a safe default with a large ecosystem. WinUI 3 fits products that want the current Fluent look and modern Windows integration, provided its controls and deployment model meet your needs. If macOS or Linux is likely during the product’s lifetime, I’d evaluate Qt, Avalonia, or a shared core with native shells before the first screen is written.

Question 3

Can you modernize a legacy WinForms, MFC or Swing application without rewriting it?

Accepted Answer

Usually, and it’s often the lower-risk path. Typical steps are moving to a supported runtime (modern .NET or a current LTS JDK), fixing high-DPI and threading problems, putting tests around the business logic, and then replacing screens or modules one at a time. A full rewrite only makes sense when the existing code blocks something the business needs and incremental change can’t get there.

Question 4

How low can video latency go?

Accepted Answer

That depends on the network, the encoder, the transport and the display, so I don’t quote a number before measuring. As a rough guide: on a LAN with tuned encoder settings, well under a second glass-to-glass is realistic; WebRTC can stay under a second across the internet; SRT adds a latency window sized to the link; LL-HLS sits in the low seconds. I agree a budget with you, then measure against it on your hardware and network.

Question 5

Do you use FFmpeg and GStreamer, and what about licensing?

Accepted Answer

Yes, both, chosen per project. FFmpeg’s libraries are LGPL by default, but enabling GPL components such as libx264 makes the build GPL, and codecs like H.264 and HEVC may also involve patent licensing through patent pools. GStreamer’s core is LGPL, while its plugins come under a range of licenses. I’m not a lawyer and don’t give legal advice; I document every component, build flag and codec in the product so your counsel can make the call.

Question 6

Can you integrate our cameras or hardware?

Accepted Answer

In most cases, yes. For IP cameras I use ONVIF and RTSP and test against the models and firmware you actually deploy. For instruments and controllers I work with serial, USB, Modbus and TCP/UDP protocols, or wrap the vendor SDK. I need access to the hardware — loaned units, remote access to a test bench, or at minimum protocol documentation and traffic captures — and I usually build a simulator early so development doesn’t depend on the bench.

Question 7

Which platforms do you build for?

Accepted Answer

Windows, macOS and Linux. Windows work is usually C# and .NET or C++; macOS and Linux work is usually C++ with Qt, Java, or a cross-platform .NET UI over a shared core. Shipping on macOS requires an Apple Developer account in your organization’s name for signing and notarization, and I’ll tell you early what each platform needs from you.

Question 8

How do you handle installers, code signing and updates?

Accepted Answer

As part of the project, not an afterthought. I set up MSIX or WiX installers for Windows, signed and notarized packages for macOS, and deb, rpm, AppImage or Flatpak for Linux, all produced by CI. Code-signing certificates should be issued to your organization; I configure the pipeline to use them through a hardware token, HSM or cloud signing service. Updates use signed manifests and staged rollout, so a bad release can be stopped before it reaches everyone.

Question 9

Our app crashes on some customers’ machines and we can’t reproduce it. Can you help?

Accepted Answer

Yes. I add crash reporting that captures minidumps for native crashes and full exception details for managed code, set up a symbol server so every report resolves to source lines, and add hang detection for the UI thread. With a few real reports in hand, most field crashes become diagnosable: the dump usually points to a race condition, a use-after-free, a driver problem or a device state nobody handled.

Question 10

How do you test desktop and video software?

Accepted Answer

In layers. Core logic gets unit tests with xUnit, Google Test or JUnit. Media pipelines get tests with reference clips and synthetic sources that check frame counts, timestamps and A/V sync. Device code runs against simulators and replayed captures, with hardware-in-the-loop runs before release. The UI gets automated smoke tests through the platform accessibility APIs, plus manual checks on high-DPI, multi-monitor and screen-reader setups. Long-running apps get soak tests measured in days, not minutes.

Question 11

What do we get at the end of a project?

Accepted Answer

The code lives in your repository from the first commit, with build scripts, CI pipelines and architecture notes alongside it, so another engineer can pick it up. I also write down the decisions that would be expensive to rediscover: threading rules, protocol details, encoder settings and the reasons behind them.