I have a collection of fully-annotated, low-resolution (<720 p) eye photographs and need a working implementation of Model-Agnostic Meta-Learning (MAML) that can predict the presence or absence of anemia from a single image. You will receive the images, their labels, and a straightforward train/validation/test split. I am looking for a clean, well-commented Python pipeline—preferably in PyTorch, though TensorFlow with Keras is acceptable—that (1) prepares the data, (2) trains a MAML backbone on the meta-task, and (3) evaluates performance on the held-out set. Because the photos are small, any sensible preprocessing or augmentation you recommend is welcome, provided the final model can be reproduced with deterministic settings. Deliverables: • Source code and environment file (requirements.txt or environment.yml) • Training script and instructions to run it on a single GPU • Saved best-performing weights • Brief README summarising results and how to fine-tune on new data I will consider the job complete when I can reproduce your reported accuracy from a fresh clone on my machine.