I need a Python-based AI system that takes our internal image dataset, learns to segment objects accurately, and then plugs straight into our existing web and mobile front-ends. The core must use OpenAI, LangChain, TensorFlow, PyTorch—or any comparable stack you prefer—as long as the code is clean, well-documented, and easy for my team to maintain. Scope • Automatic image segmentation is the primary function; no other vision tasks are required right now. • The model must train on our custom dataset (we’ll supply labeled images) and allow incremental retraining when new data arrives. • A lightweight REST or GraphQL layer should expose the inference endpoint so the website/app can call it in real time. • All processing and automation logic must be written in Python; feel free to introduce supporting libraries (FastAPI, Celery, etc.) if they improve performance. • End-to-end delivery needs to land within two to three weeks, including a quick hand-off session. Deliverables 1. Training pipeline with reproducible scripts or notebooks 2. Inference API fully integrated and live on our staging environment 3. Source code repository with clear README and environment files 4. Brief technical walkthrough (video call or recorded demo) Acceptance criteria – Mean IoU on our validation set meets the target we’ll agree on before training starts – API returns segmentation masks in <500 ms on standard GPU or optimized CPU instance – Zero hard-coded credentials; environment variables or vault-based secrets only – Code passes pylint/flake8 and basic unit tests supplied in the repo Please include links or repos demonstrating previous AI/ML projects—especially anything involving image data—and outline how you’d hit the timeline.