Blog
Writing on platform design, delivery choices, security decisions, and product-building patterns.
This project shows how I turn noisy infrastructure security findings into an explainable, remediation-focused workflow by combining proven scanners with bounded agent reasoning.
The difference between an AI demo and an AI product is everything around the model: UX, retrieval quality, runtime setup, packaging, trust, and documentation.
AI PDF Reader brings together product architecture, privacy, retrieval quality, and delivery discipline in one local-first system.
Local-first AI products feel safer by default, but they still need careful threat modeling around files, models, prompts, packaging, and desktop boundaries.