I need to embed two core AI capabilities into my current workflow: a predictive-analytics engine driven by machine-learning techniques and an image-recognition module powered by computer vision. Both tools should be production-ready, well-documented, and callable through REST or Python APIs so they slot cleanly into our existing stack. The predictive component will ingest historical data that I will provide (CSV and SQL sources) and return forward-looking metrics such as demand forecasts and risk scores. Accuracy must be benchmarked against a held-out test set, with clear reporting on feature importance and model performance (precision, recall, F1). The image-recognition piece will classify and tag uploaded photos, surfacing confidence values for each label. A lightweight front-end demo or notebook that shows the model running on sample images will help us validate results quickly. Preferred toolchain includes Python, TensorFlow or PyTorch for model development, plus standard libraries for data handling (Pandas, scikit-learn, OpenCV as needed). If you have a different stack that still meets these requirements, mention it and explain the trade-offs. Deliverables • Trained predictive-analytics model with reproducible training script • Trained image-recognition model with inference endpoint or callable function • Clean, commented codebase stored in a private Git repo • Setup guide and brief technical documentation (max 5 pages) Acceptance criteria • Predictive model hits agreed accuracy thresholds on unseen data • Image model returns ≥90 % top-3 precision on validation set • Code passes linting tests and installs via one-command setup Timeline and milestone suggestions are welcome once you’ve reviewed the scope.