End-to-end Sinhala Sign Language recognition pipeline combining video annotation, pose estimation, and mobile deployment. Raw SSL video datasets flow through VoTT annotation and OpenCV clipping, then Google MediaPipe Holistic extracts landmark coordinates for sequence modeling. An LSTM neural network classifies sign sequences, which are optimized via TFLite conversion and quantization for deployment on a Kotlin-based Android app that outputs recognized Sinhala text. This architecture demonstrates best practices for real-time gesture recognition on mobile devices with minimal latency and resource constraints. Fork and customize this diagram to adapt the pipeline for other sign languages or gesture recognition tasks on Diagrams.so.