Projects

Mlx-cluster (Open Source)

A Python library for generating and calculating random graphs efficiently on Apple GPU using mlx-graphs and mlx.

Utilized Metal kernels with Python for efficiently calculating random walks and biased random walks, achieving 1.2x performance compared to regular PyTorch kernels on macOS
Used Nanobind, Poetry, and Python Packaging Index (PyPI) for publishing the package

Technologies: Python, MLX, Metal, Nanobind, Poetry

Created a malware detection algorithm that can detect malware by disassembling opcode sequences.

Utilized BERT and XLNet architectures and various ensemble methods to improve accuracy from 95% to 97%
Novel approach using NLP techniques for cybersecurity applications

Technologies: Python, PyTorch, BERT, XLNet, Transformers

Utilized graph machine learning algorithms to classify illicit nodes in bitcoin transactions

Utilized GraphSage, GNN, GCN to classify illicit nodes
Utilized GNN explainer to verify which nodes contribute for majority of illicit nodes transaction
Contributed to make this dataset available in mlx-graphs

Contributed to mlx-graphs library ranging from datasets to graph transformations

Technologies: mlx