I need a production-ready object detection model built, trained, and packaged so it runs smoothly on iOS and Android devices, in a modern web browser, and as a lightweight desktop application. The same model should power every platform to keep accuracy and behaviour consistent. You are free to choose the framework you are most comfortable with—TensorFlow, PyTorch, YOLOv8, Detectron2, or another proven library—as long as the final artefacts meet these requirements: • Mobile: optimised builds (e.g. TensorFlow Lite, Core ML, or ONNX) that hit realtime speeds on mid-range phones. • Web: WebAssembly/WebGL or TensorFlow.js implementation that loads in under three seconds on a standard connection. • Desktop: a small executable or Python app with GPU support when available and CPU fallback when not. Deliverables 1. Trained model files and all training scripts. 2. Inference wrappers for each target platform. 3. Step-by-step build & deployment instructions so I can reproduce everything locally. 4. A simple demo for each platform proving the model detects objects accurately. I will supply the labelled data and clarify the target object classes once we start. Accuracy, speed, and reproducibility will be the acceptance criteria. If this scope matches your expertise, let’s discuss timelines and any additional resources you might need.