Technical Writing

The design and philosophy behind CiteMentor's micro-royalty system — tracking attribution down to the paragraph and building an economic layer for AI-assisted reading.

A practical pipeline for deduplicating and organizing large image datasets before training — perceptual hashing, clustering, and automated labeling at scale.

Practical patterns for writing faster, cleaner Python — profiling hot paths, vectorization, generator pipelines, and idioms that compound across a large codebase.

Part 2: connecting flow-based generative models to modern TTS architectures — Glow-TTS, VITS, and how normalizing flows enable parallel, non-autoregressive synthesis.

A ground-up explanation of normalizing flows and how they enable high-fidelity neural text-to-speech synthesis — from invertible transformations to tractable likelihoods.