At FlyRank AI, I completed a Google Search Ranking & Discoverability capstone for content-opportunity scoring using approximately 30,000 Google Search Console/GA4 records. After identifying an overly optimistic initial evaluation, I used client-grouped train/test splits to reduce data leakage and reported an out-of-sample Precision@10 of 0.40.
For my BRAC University undergraduate thesis, NeuroVentura, I developed a late-fusion framework combining a frozen MONAI 3D DenseNet-121 MRI, XGBoost on MMSE, age, and gender, and a residual MLP on a matched cohort.
My projects include a deep learning pipeline for lung cancer subtype classification and a four-class diabetic retinopathy severity classifier using ResNet50 transfer learning. I also compared LSTM autoencoder, VAE, and Transformer approaches for music generation.
I’ve developed MERN applications for community issue reporting and academic resources, as well as a PHP healthcare platform. As a freelance tech writer and researcher, I produce research-based academic and technology articles with supporting visual content.

