AI as a working layer, not a talking point.

Sovereign AI is being announced faster than it is being understood. Leaders need a map, not another news feed.

A structured intelligence system with country-level records, policy tracking, events, research, and signal feeds. 130+ programs across 50+ countries.

Classifying sources, summarising policy documents, extracting entities, identifying focus areas, enriching records, and turning raw signal into structured intelligence.

In active build. Live at sovereignaivector.com with continuous editorial and pipeline work.

Most team LinkedIn profiles quietly drift out of sync with the team's brand and the story the org is trying to tell. Inconsistent headlines, outdated employers, and the occasional departure signal — all visible to any buyer who looks before a meeting.

A four-tool workflow that turns a list of LinkedIn URLs into a ranked team scoreboard with individual assessment tabs. Five criteria, scored one to three, total out of fifteen, with status labels that point to a specific next move.

Enrichment via Clay. Structured staging in Sheets. Claude reads the structured record and the live profile, scores across all five criteria, and returns a JSON payload. Apps Script rolls each payload into a named tab and updates the team scoreboard.

Working framework. Tested on a real team; the methodology applies to any LinkedIn assessment use case.

Data center infrastructure is mostly invisible to the people deciding to invest in it. Modular builds in particular get described in stats — megawatts, PUE, deployment speed — without the texture of what is actually inside one.

An interactive VR walk-through that loads in any modern browser. A sequence of rooms, each with hotspots that open short, plain-English explanations. Self-paced for a microsite, and kiosk-friendly with an idle reset.

AI as co-builder. Drafting scene structures, scaffolding the 3D code, generating hotspot copy, and pressure-testing the sequence. The judgement about what to show, what to leave out, and how to name each room was the human work.

Working prototype. Runs on a laptop or a touchscreen kiosk.

M&A teams often start with a long list from a corporate-finance partner, then spend weeks turning that list into a shortlist they can actually act on. The work is structure, evidence, and consistent screening across every name.

A static, single-page dashboard with four tabs — Overview, Master List, Company Detail, Evidence & Audit — built around a two-stage screening method: structural priority first, then strategic recommendation. Self-contained, deployable behind access controls.

AI-enriched research per target: ownership verification, vendor partnerships, business description, ecosystem relevance, and diligence risk flags. The structured outputs feed every tab and every KPI.

Working prototype, used in a confidential engagement. Industry-agnostic; the screening framework is the reusable part.

A decade and a half of clear operating thinking sits as long-form posts. Operators rarely have time to read three thousand words to find the one decision lever they actually wanted.

A static microsite with two volumes and an appendix. Vol. 01 Marketing, Vol. 02 SaaS Metrics — Rule of 40, the Mendoza Line, the Magic Number, NRR, and the burn multiple, each as a calculator rather than a chart — plus a curated reading list. Every chapter cites the underlying post.

AI as research and authoring partner. Reading and clustering posts, drafting chapter summaries, designing the interactive calculators, and writing micro-copy. The judgement about what to surface, how to frame it, and what to leave out was the human work — and the author's review.

Private staging build, prepared for the author's review prior to public launch.

Partner teams operate inside a tangle of vendors, partners, programs, customers, and internal roles. The traditional partner-management playbook scales linearly. The complexity scales exponentially.

A conceptual data model and capability layer for a partner-ecosystem copilot. Five capabilities — personalised journeys, real-time guidance, predictive analytics, enhanced collaboration, and ontology-powered intelligence — sitting on a stack from infrastructure up to the partner-facing interface.

Two-sided. PartnerGrok is itself an AI proposal — the case for what an AI copilot in this domain should actually do. AI also helped sharpen the framing, the data model, and the language that channel chiefs would recognise as their own.

Published concept paper. Open to a build conversation with the right team.