AI, where it earns its place.

Products where AI is the core of the value, not a bolt-on. Agentic systems, LLM pipelines, retrieval and search over messy real-world data. The engineering that makes them dependable: evaluation, guardrails, observability and cost control.

The proof: at S-RM I architected a platform streaming millions of documents a month through NLP grading, so analysts read the ones that matter. Read the case study →

The web is getting a second kind of visitor. AI agents now read sites, answer questions from them, and act on them for their users. A site that is easy for agents to read gets cited, quoted and recommended; one that isn't becomes invisible to a growing share of the audience.

This site is the demonstration. It follows the emerging agent-readiness standards end to end: a curated llms.txt index, markdown versions of every core page (ask for text/markdown and you get it), content signals telling AI crawlers what they may do, published agent skills and a machine-readable API catalogue. Run it through Is It Agent Ready yourself, or read the standards at The Website Specification.

I audit and retrofit existing sites and applications the same way.

Most teams don't need an AI product. They need AI applied carefully inside the business they already run: agentic search across research libraries and document-heavy systems, so the answer surfaces instead of the afternoon disappearing. AI-assisted engineering that ships faster without shipping worse, because a senior engineer stays accountable for every line. And an honest map of where AI does not belong yet.