BEGIN A CONVERSATION · TAKE THE FREE MAO INDEX · READ THE ESSAYS · LINKEDIN
I worked at OpenAI, led AI for product experience at Spotify for a platform serving 675 million people, advised Coinbase, and helped build infrastructure at Stripe. I'm based in Dubai and work across North America, Europe, Asia-Pacific, and the Gulf.
What follows is not a portfolio of demos. Each repository is a working, tested answer to a problem that organizations are paying real money to get wrong. Find the problem you recognize.
A private, interactive demonstration. Choose a global bank, a sovereign program or a consumer AI company, break the system, and watch five of these tools diagnose a simulated production failure. The inputs are synthetic and labelled as such. Every analytical result is the return value of the real code in this profile, running in your browser.
Enter the Control Room · The Receipts · The Million Dollar Question
 |
 |
 |
 |
Most AI programs fail in the gaps between the model, the work it is supposed to do, and the system that coordinates both. mao-framework makes those three layers explicit and testable: policy gates, verified steps, spend budgets, and a hash-chained audit log, in code anyone can read in an afternoon. It also includes a twelve-question diagnostic that finds the layer where your system will fail first.
Want the diagnostic without the code? Take the free MAO Index.
For leaders allocating real budget to AI infrastructure. Read the thinking first: The Meter Is Running. Nobody Built the Architecture to Stop It.
| Repository | The question it answers |
|---|---|
| model.ledger | Which team and which model is driving the bill, and when did it spike? |
| token.econ | Does this product make money per user, and what price hits the margin we want? |
| roi.lens | What is the ROI of this use case as a range with odds, not one hopeful number? |
| build.buy | Build or buy, over five years, at our cost of capital? |
| route.sense | Which model actually earns its cost on our traffic? |
For organizations running systems whose failures have consequences. Read the thinking first: Nobody Signs for the Agent
| Repository | The question it answers |
|---|---|
| eval.forge | Did this change make the system worse, and can we block the release? |
| drift.watch | Has the world moved away from what the model was built on? |
| guard.rails | What is leaking out, and what is being injected in? |
| prompt.vault | Who changed the prompt, what changed, and can we roll it back? |
| incident.log | How severe is this failure, who owns it, and did we miss the SLA? |
| agent.trace | Where did the agent run spend its time and money? |
| lex.machine | Which EU AI Act, NIST AI RMF, and GDPR controls do we have evidence for? |
For leaders deciding what to build, what to buy, and where to invest. Read the thinking first: The Job Nobody Has Yet
| Repository | The question it answers |
|---|---|
| readiness.index | How ready are we, which gap matters most, and in what order do we close them? |
| vendor.score | Does the winning vendor survive a change in our priorities? |
| orchestra.dsl | What will this AI workflow cost, how long will it take, and where does a human approve? |
| context.engine | What belongs in the context window when the budget is fixed? |
| playbook.compiler | Are our operating playbooks complete, numbered, owned, and checkable? |
For governments, regional institutions, and businesses operating outside the assumptions built into American AI infrastructure. Read the thinking first: Your AI Has a Passport Problem and The Language They Did Not Build For
| Repository | The question it answers |
|---|---|
| sovereign.map | Which of our AI data flows leave the jurisdictions our rules allow? |
| lisan | How do we normalize Arabic text so search and models actually match it? |
| nusantara | How do we handle rupiah, Indonesian dates, phone numbers, and everyday slang in AI pipelines? |
Every repository above is dependency-free Python with a test suite and continuous integration. They are my own work and were not built for any client.
Press. The San Francisco Standard, July 2026: Can't get an AI job? Hire someone who already has one. The article describes me as a former product strategist at OpenAI and head of AI at Spotify.
Conversations. What OpenAI, Spotify, and Coinbase Taught Me About Building AI People Actually Use (Celia's Corner). Why Your AI Strategy Is Burning Cash (And How to Fix It) (KAJ Business and Money Live).
Writing. blog.andmaverick.com on AI economics, accountability, sovereignty, and language.
Advisory. Selected organizations, as published at andmaverick.com: Government of Australia, Government of New Zealand, Blackstone, Mayo Clinic, JPMorgan Chase, Microsoft, Suno, Temasek, G42, Sea Group.
Fiction. I also write. The Looming Horizon is my dystopian novel, with a companion series on YouTube.
AndMaverick works with sovereign wealth funds, governments, frontier AI companies, and enterprises. There is no team to hand you to. I lead each mandate personally.
If one of the problems above is yours, the next step is a conversation.
REQUEST A PRIVATE ARCHITECTURE REVIEW
BEGIN A CONVERSATION · TAKE THE FREE MAO INDEX · INQUIRE ABOUT SPEAKING
Klarheit aus Chaos.