grep -rl "abstract" ~/research
Papers from my M.S. in Computer Science at LIU Brooklyn, including IEEE-format research and technical reviews.
99.2%external validation accuracy
An end-to-end hybrid framework for automated, non-invasive Parkinson's Disease detection from digitized hand-drawn graphomotor biomarkers. It combines deep feature extraction from multiple CNNs (ResNet50, VGG19, InceptionV3), Genetic Algorithm feature selection, XGBoost classification, and SHAP interpretability.
SBERTsemantic ranking over a rich user profile
A content-based recommendation system that uses Sentence-BERT embeddings and a rich user profile built from explicit preferences, liked articles and reading history. Cosine similarity over all-MiniLM-L6-v2 vectors drives personalized news ranking, avoiding the limits of keyword matching.
Compares the two dominant deep learning frameworks across developer experience, training and inference efficiency, deployment, and ecosystem. Both deliver strong results, but PyTorch leans toward simplicity and research while TensorFlow leans toward a complete production ecosystem, so the right choice depends on which trade-offs matter most.
A structured review of the TensorFlow paper: its single dataflow-graph model for computation and mutable state, extensibility through user-level libraries, distributed training, and benchmark performance on image classification and language modeling across CPUs, GPUs and TPUs.
A qualitative analysis of microservices and distributed systems through case studies of Netflix, Uber and ChatGPT, showing how system analysis and design principles (modularity, scalability and fault tolerance) address the limits of monolithic architectures at production scale.