Side-quests.

A self-contained benchmark measuring how much an NVIDIA GPU accelerates the core operation behind deep learning, large dense matrix multiplication, compared to the CPU. Tests four numeric precisions (FP32, TF32, FP16, BF16) to expose the speed-versus-accuracy trade-off that mixed-precision training relies on. Up to 59x speedup on an RTX 3060 Laptop.

Three Python tools behind one CLI covering the password lifecycle: entropy audit with breach-check via HIBP k-anonymity, ranked hash identification, and dictionary-attack driver over John the Ripper and Hashcat. Standard library only, no third-party crypto.

Two-part Streamlit project on Maricopa County, Arizona residential property sales using public Assessor data. An interactive dashboard for exploration, plus an end-to-end ML pipeline that predicts sale price (Random Forest: 0.846 R², 13.14% MAPE on 91k test set).

Multi-threaded TCP port scanner and service fingerprinter, written in pure Python with no third-party dependencies. A hand-built alternative to nmap -sT -sV, built to understand what a port scanner actually does under the hood.

The same handwritten-digit classifier built four ways, from sklearn logistic regression (92%) through a NumPy from-scratch neural net (97%) to a PyTorch CNN (99%+). Deployed as a Streamlit 'draw a digit' demo.