I need a unified AI solution that can both analyse images and act as a lightweight personal assistant for my inbox. The primary requirement is a robust image-recognition or analysis pipeline: think accurate classification, object or feature detection, and an easy way for me to feed new images in and retrieve clean, structured results out. I’m open to using TensorFlow, PyTorch, or a comparable framework as long as the final model is maintainable and can be retrained when I add fresh data. Alongside that I’d like a slim language-model layer focused on personal-assistant tasks—specifically managing emails and messages. Automatic triage, smart labelling, concise reply suggestions and quick summaries are the key features I’m after; no broader customer-service or scheduling tools are needed right now. Deliverables • Trained image-recognition model with evaluation report • NLP component that connects to an IMAP / Gmail-style mailbox, ranks importance, drafts replies and returns JSON output • Clean, well-commented source code plus a README showing setup, inference and retraining steps • Brief video or live demo proving both parts work together (image analysis and email assistant) I’ll supply sample images and a test mailbox as soon as we kick off, and I’m happy to iterate fast—clear metrics and reproducible results matter more than flashy UI.