[BhavyaQuantDeveloper]

This PR adds my GSoC 2026 final blog post about MedCapsNet.jl, a native Julia/Flux.jl reproduction of Capsule Networks against Medical Imaging Data Challenges (Jiménez-Sánchez, Albarqouni & Mateus, LABELS@MICCAI 2018).

The post covers:

• The CapsNet architecture (dynamic routing, margin loss) and the LeNet / Baseline-CNN comparisons
• The limited-data / class-imbalance / augmentation experiment harness
• Results in the low-data regime and a clearly-attributed comparison against the paper's published F1 tables
• Findings from the reimplementation itself (a Macenko stain-separation bug caught by tests, a flaky-seed reproducibility bug)
• The medical preprocessing pipelines (TUPAC16, DIARETDB1) and an Apple-GPU (Metal.jl) porting postscript

All figures are outputs of the project's own Julia code.

• Code: https://github.com/BhavyaQuantDeveloper/MedCapsNet
• Project page: https://bhavyaquantdeveloper.github.io/MedCapsNet/

New files under JuliaHealthBlog/posts/bhavya-gsoc-2026/ only — no existing content touched.