Machine-intelligence work backed by more than two decades of software delivery
Michael architects software around agents, models, memory, retrieval, validation, recovery, and human review. His public work emphasizes explicit boundaries: identity and scope for cooperating agents, durable context with provenance, observable runtime behavior, and governed integration with established business systems.
The engineering foundation spans more than twenty years of .NET, Microsoft Azure, SQL Server, TypeScript and Angular, automated testing, legacy modernization, architecture, and technical mentoring.
From model capability to inspectable software systems
Michael’s current focus includes multi-agent coordination, persistent AI memory, local model and runtime components, semantic retrieval, structured outputs, source governance, and bounded automation. The goal is not to treat an AI model as an unbounded authority, but to place it inside software with clear contracts, diagnostics, recovery paths, and human approval.
Why this subject connects to that body of work
MachineTradecraft.com examines the difference between what a person sees and what a parser, tokenizer, retriever, model, or agent receives. That perception gap is directly relevant to trustworthy AI ingestion, prompt-attachment handling, canonicalization, provenance, and privilege separation.
The site turns those concerns into dependency-free guides, bounded laboratories, benign fixtures, and defensive research that can be inspected locally without analytics, remote scripts, or third-party runtime services.