AI solutions architecture and advisory. I evaluate, design, and build LLM-based systems — and tell you when you shouldn’t.

Twenty years of software systems architecture. Two years designing production AI. Independent assessment, architecture, and technical talks for companies deciding what to do about AI.

I spent two decades designing software systems in logistics, healthcare, ecommerce, and financial services — routing engines, real-time prediction systems, rules engines, ecommerce platforms. I began working with LLMs approximately two years ago, as an architect rather than a researcher, evaluating where the technology fits inside production software.

The conclusion I reached: model selection is rarely the constraint. Integration architecture, evaluation methodology, and error handling are. A system operating at 95% accuracy produces incorrect output 5% of the time, and most implementation failures trace to teams that never designed for that.

I assess AI proposals, design the systems, build the proofs of concept, and advise the teams. When the honest answer is that AI is the wrong tool for a given problem, that is the answer I give.