Fluency is not accuracy
AI systems fail quietly and in complete sentences. I help companies put them into production anyway, with the evidence, checks and structure that make the output something you can act on.
Advisory
Implementation and product work for companies putting AI into something people depend on. Retrieval and context design, evaluation before deployment, and the unglamorous decisions about what happens when the model is uncertain. Engagements start with a call, not a proposal.
Why AI Gets It Wrong
A short book on how language models produce language, why that mechanism makes certain failures inevitable, and what actually reduces them. No prompt templates. No mathematics. Readable in an evening.
Writing
Working notes on evaluation, retrieval, memory, tools and agents. The material that separates a demonstration from a system people can safely depend on.
Where this comes from
Over a decade of product work across SaaS, insurtech, e-commerce and data intelligence, in Bahrain, India and Canada. The book is the argument. This is the background it came from.
The reader list
New writing on building AI systems that hold up in production, sent when there is something worth sending. No cadence, no filler.