Aria Mohajer
The most valuable knowledge in any expert domain was never written down. That is not a gap in the dataset. It is a gap in how we have thought about intelligence.
Every AI system built today learns from what humans have written down. That is also its ceiling. The knowledge that defines true expertise does not exist in any dataset. It lives in the people who built it over decades, and when those people retire, it disappears.
DeepHavn is building the infrastructure to encode that knowledge and create intelligence systems that learn from outcomes. Not labels. Not human annotation. Outcomes. The system improves from that signal permanently, with no ceiling imposed by human accuracy.
Finance is where we prove it first, because the outcomes are binary and the cost of getting them wrong is catastrophic. From there, the mechanism applies wherever expertise is implicit and stakes are high.
Tacit Intelligence Research Lab. We encode the knowledge that has never made it into any dataset and build systems that learn from outcomes. First product: AICIL. Delaware C-Corp, headquartered in San Francisco.
Curated ecosystem connecting founders with capital and the right partners. Bridging emerging technology into regulated industries across the Gulf.
Gig economy platform reaching six figures in revenue before COVID-19 destroyed the physical gig market in 2020. I did not find a pivot. I shut it down cleanly and studied the failure carefully before moving on.
$200 trillion moves across borders every year. $30 trillion of it is delayed, suspended, blocked, or returned. Not because the money is bad. Because no system can reliably predict whether a payment will satisfy the compliance requirements of every correspondent bank in its chain before it is sent.
The current approach is supervised learning on labeled transaction data, reviewed by human compliance teams. The system is bounded by the accuracy of those humans. That sounds acceptable until you apply it to $200 trillion.
AICIL predicts clearance before a transaction is sent. Every outcome is a training signal. The system does not plateau. Finance is where the mechanism is proven. The research agenda runs further.
My father came to Germany at twelve. He left during the Gulf War, alone, with nothing. My mother came at seventeen and studied medicine while raising me. I started my first company at fourteen. A custom suits brand, manufacturing in the UK and Portugal. Not a school project. A business.
I am 26. Iranian-Kurdish, born and raised in Germany, now based in Dubai. I have been building things since before I understood what building meant. The understanding came later. The compulsion was always there.
I think about what my parents gave up to give me the platform I have. Not as inspiration in the abstract. As a specific, daily obligation. The work has to be worth the sacrifice.
I withdrew from university to build DeepHavn full-time. That was not the hard decision. The hard decisions were the earlier ones: learning to think precisely enough to see a problem nobody else had framed clearly, and caring enough about the answer to go after it.
I believe the path to intelligence that surpasses human experts runs through tacit knowledge and outcome-based learning. Not through more labeled data and larger models. I believe financial infrastructure is one of the most consequential and least understood systems on earth. And I believe most people building AI right now are optimizing in the wrong direction.
These are not opinions I hold lightly. I have spent a long time thinking about them. DeepHavn is the result.
90 million people. 98% youth literacy. A scientific output that ranks in the top 20 globally, produced under the most severe sanctions regime in modern history.
A diaspora that built Uber, Databricks, eBay, and Intuit. Trillions of dollars in collective value, created entirely outside the country by people who had to leave to build anything at all. Before 1979, more than 90% of Iranian students who went abroad came home. Today, fewer than 10% do.
Most people looking at Iran today are looking at the last 47 years. The actual frame is 7,000. Iran produced the world's first declaration of human rights, the first postal system, and the first professional legal code. Civilizations that have built extraordinary things once tend to do it again when the constraints suppressing them are lifted. The constraint here is governance. It is changing.
The angle almost nobody is modeling: AI's binding constraint right now is energy. Every major lab is scrambling for power. Iran has the world's second-largest natural gas reserves, 300 days of sun per year, and 90% of its land mass suitable for solar generation at some of the highest radiation levels on earth. The talent has always been there. The capital is in the diaspora. The energy is in the ground and the sky.
The South Korea of 1960 had a GDP per capita of $158. Today it is a top-12 economy. The variable that changed was governance. When the conditions change for Iran, the compounding will not be linear.
I am not a politician and I do not spend my time on this. But I think clearly about large systems, and this is one where the numbers are too significant to ignore.
National blockchain study produced in cooperation with Statista and the German Federal Ministry for Economic Affairs and Climate Action. Engaged 50+ German companies for data collection and advised on forming a national blockchain ecosystem. Study presented to the German Parliament. View study →
Built during the Web3 wave. Product strategy, five-quarter roadmap, dev, design, and community teams. An exercise in building organizations fast under market conditions that do not wait.
Global retail data analysis and supply chain coordination.
Custom suits brand with production in the UK and Portugal. Supplier negotiations, brand identity, first collection. Most 14-year-olds were doing something else.
If what you read here is interesting to you, I want to hear from you.