I have an existing project already running on Google’s MediaPipe Hand Landmarker demo, and I now need to extend it. Your first objective is to create new functionality on top of the current pipeline—think fresh gestures, custom landmarks, or richer event callbacks that let the application react in real time. While doing so, I also want you to tighten up finger-position accuracy; tiny mis-reads in joint locations are still slipping through and affect downstream logic. Access to the source will be granted once we have a signed NDA. Please respond with a short note on similar work you have done with MediaPipe or comparable libraries so I can gauge fit quickly.