Freelance data scientist · All case studies

6,270 animals. 151 types. One classifier.

A deep-learning challenge on a wide, fine-grained animal set. I built torchvision architectures, tuned the training recipe, and pushed accuracy without wasting the budget on a bloated net.

Client: Vision challenge. Built by Dilshad Raza.

One hundred and fifty-one animal types on a few thousand images is not ImageNet. Similar coats, similar poses, and a long tail of rare classes — the model had to generalise, not memorise the training folder.

I developed image-classification models in torchvision, swept hyperparameters, and applied data augmentations aimed at pose and colour shift. The target was high accuracy with an efficient footprint — a model you can actually train again.

Hire Dilshad Raza for similar freelance data science, machine learning, and AI automation in the UK, United States, and Australia.