Alexandre Elkrief and the machinery of trust

The interesting thing about Alexandre Elkrief’s crypto career is that it contains a newsroom. Before the smart contracts, before the lending infrastructure, before August and Upshift, there was Knowhere, a company trying to help readers make sense of competing accounts of reality. News and finance have an uncomfortable habit in common: both ask people to trust an account of events they cannot entirely inspect. Elkrief has worked on the machinery behind both.

In 2018, he described Knowhere’s ambition with a short, hopeful phrase: “adding data to journalism will bring back trust.” It was a considerable assignment for a dataset. The company was developing machine learning and text understanding tools for news, with Elkrief as co-founder and president. Its other co-founders were Nathaniel Barling and Dylan Rhodes. A reader might have expected the usual argument about which publication deserved belief. Knowhere wanted to examine how information became an article in the first place.

That makes a useful starting point for understanding the co-founder and co-CEO of August Digital. His professional life crosses mathematics, machine learning, media and decentralized finance. The industries change. A recurring problem remains visible: useful information gets scattered, expert judgments become difficult to inspect, and the person at the other end receives a deceptively tidy result. Connecting those episodes is an interpretation of his work, rather than a claim that he planned every turn from the beginning.

A bank analyst meets a classifier

Elkrief studied mathematics at Imperial College London, then computational and mathematical engineering at Stanford. His early résumé also includes work at Goldman Sachs and Deutsche Bank. There is a more revealing artifact than the familiar institution names: a Stanford machine learning project dated December 11, 2015, co-authored with Thomas Ayoul and Nicolas Meiller, with Youssef Ahres as mentor.

The project examined whether bank analysts’ recommendations could help predict movements in Eurostoxx 50 stocks. The team built a long-short trading strategy around those predictions, back-tested it and explored different classifiers. Logistic regression, support vector machines and random forests all entered the investigation, along with trend-following and mean-reversion features. For anyone hoping that financial judgment could be reduced to one very impressive spreadsheet, the list of experiments offers a useful corrective.

This was coursework, not a record of investment performance. Its interest lies in the question being asked. An analyst produces an opinion. A model tries to discover whether that opinion contains a usable signal. Years before Elkrief was explaining on-chain trading infrastructure, he was already examining the distance between an authoritative statement and something that could be tested. The paper supplies a concrete early example of his interest in applying computation to financial information.

The algorithm still needed an editor

Knowhere was founded in 2015 and went through Stanford’s StartX accelerator. By the end of 2020, its audience was listed at 540,000 and its funding at more than $7 million. Those figures describe the company at that point in its life. They also give the earlier media chapter some scale: this was a working publishing business, with readers and investors, rather than a passing experiment in automated prose.

Its process began by collecting and comparing coverage. The system helped identify which stories were attracting attention, then organized information for writing. Elkrief described a structure that mapped facts and their relationships. The ambition was to find a coherent route through a large body of material. Anyone who has opened sixteen browser tabs to understand one event can appreciate the appeal, even before the software gets involved.

Human judgment remained inside the operation. Editors reviewed story priorities and checked drafts; Barling described people and machines as partners with different strengths. The machine could examine large quantities of material, while editorial decisions still needed people. That detail matters to Elkrief’s story. His first company’s answer to an information problem included both automation and a human decision layer. Technology supplied a way to organize the work, while responsibility still had to sit somewhere.

“adding data to journalism will bring back trust.”Alexandre Elkrief, 2018

A different kind of exposure

Elkrief subsequently led decentralized finance at LedgerPrime, a digital asset investment firm. That role placed him on the investment side of the systems he would later help build. In an April 2025 public post, he described joining DeFi about five years earlier, attracted by the possibility of broader access to wealth creation. He also described the complications that had accumulated around that idea: difficult products, conflicting incentives and opaque risk management.

His partnership with Aya Kantorovich brought another institutional perspective. She had been on the founding team at FalconX; he had worked on DeFi strategies at LedgerPrime. Their company, initially called Fractal and later renamed August, focused on infrastructure for institutions to clear, settle and margin digital asset trades on-chain. The vocabulary is forbidding. The underlying questions are more familiar: where are the assets, what supports the loan, and who can see a problem developing?

In April 2023, Fractal announced a $6 million seed round led by Hack VC. The co-founders’ stated priorities included visible reserves, collateral management involving multiple parties and on-chain settlement. Those are specific operating ambitions. A lender needs a view of the collateral supporting an obligation. A trading firm needs to understand exposure across its positions. Putting transactions on a blockchain provides records; arranging those records into a usable financial operation is the business the founders chose.

Seven billion, with the date attached

August’s March 2025 funding announcement put numbers around that work: a $10 million Series A led by Dragonfly Ventures, more than $7 billion in monthly transaction volume and over $400 million in loans originated during the preceding year. The company described support for more than 12 blockchains, 250 tokens and 70 protocols. These were company-reported operating figures at the announcement date. They measure different things, and a busy trading platform’s volume should never be mistaken for its revenue.

The product joined services that traders otherwise have to coordinate across separate venues: borrowing, lending, execution, derivatives and collateral management. Its portfolio margin system was designed to let institutions use assets and positions across a broader set of obligations. The practical attraction was capital efficiency. Money committed in one place can be difficult to use elsewhere; a system that understands a portfolio has a different view from a collection of disconnected accounts.

For Elkrief, the public argument kept returning to making trust part of the infrastructure. That phrase becomes useful only when it connects to actual controls: what can be observed, what actions are permitted, and how exposure is measured. His earlier newsroom had tried to organize facts before publishing a story. August was organizing financial positions before institutions acted on them. The comparison has limits, but it explains why the career change is less random than the job titles suggest.

Nobody wants a second full-time job

Upshift added a different audience to the picture. When it emerged from stealth in April 2025, Elkrief was identified as its CEO as well as an August co-founder. The platform was being spun out of August, using its underlying infrastructure to make professionally managed DeFi strategies accessible through vaults. Elkrief’s diagnosis was unusually plain for this corner of finance: “DeFi’s super complicated.” The problem was fragmentation across chains and opportunities, with constant management becoming a job of its own.

That observation gives the product a human scale. Someone may want access to a strategy without wanting to become an expert in every network it touches. A vault packages a defined strategy into a contract-based structure, with a professional curator operating within its limits. The appeal depends on making participation easier while keeping the workings inspectable. A simple interface is pleasant; a simple interface with unexplained machinery behind it would recreate the problem the founders were trying to address.

His public writing also leaves room for a small joke at his own expense. In a 2026 LinkedIn post about changing demand for vault infrastructure, he ended with: “Probably time I added Upshift to my Linkedin.” After years of discussion about the future of financial products, the professional biography apparently needed an update. It is a welcome reminder that founders maintaining complicated systems can still fall behind on a profile field.

The paperwork arrives on-chain

The next developments were less about announcing another chain and more about fitting into existing institutions. On April 22, 2026, Securitize Fund Services announced a partnership with Upshift for independent reporting, allocation tracking and reconciliation. Professional investors need to connect activity to accounts, ownership and reporting obligations. An observable transaction is useful. An independent account of what it means for a particular investor answers another set of questions.

On July 15, 2026, Kraken Institutional announced custom Upshift vaults accessible from its qualified custody offering. Eligible clients could deploy assets into selected vault contracts and hold the receipt tokens representing their positions in segregated custody. The arrangement brought vault infrastructure into an existing institutional relationship. For the customer, that reduced the number of separate systems and providers needed to pursue a strategy. For Upshift, it put distribution alongside the technical work.

Elkrief’s stated direction for 2026 was to bring this infrastructure to traditional fintech companies and asset managers. It is a fairly practical ambition after a career spent among large abstractions: make the machinery usable inside products people already have. The former news founder now works in a business where a position, a permission and a receipt must agree. Trust is an appealing word. His career becomes more interesting when it gets down to the records, the editors and the accounts.