Martin
Markiewicz,
and the audit trail.

The pitch was never five percent. Martin Markiewicz likes to say the pitch was five percent, and then the customer engagements returned numbers that were closer to five hundred times, and he has repeated some version of that sentence in interviews for a decade. This is either founder puffery or a genuine story about the base rate of false positives inside bank compliance departments being so absurd that any reasonably competent machine-learning model, pointed at the problem, will look like a magic trick. Reasonable people disagree. Markiewicz, who trained as a mathematician, is not particularly interested in the disagreement.

Silent Eight, which he co-founded in 2013 with his sister Julia Markiewicz and Michael Wilkowski, sells software to the sort of banks that get fined a nine-figure sum every few years for missing a sanctioned entity on a wire. The product's job is to look at those wires, and the names attached to them, and decide whether an alert is worth waking a human analyst for. It does this using models trained on the bank's own history, and it produces, alongside every decision, a written narrative explaining the call. This last part is not a garnish. It is, in Markiewicz's telling, the entire reason the company exists.

The Church Street model

The headquarters sits at 3 Church Street in Singapore, a short walk from the river and closer than you would expect to the Monetary Authority of Singapore. The office is small relative to the customer list, which includes several of the largest banks in the world, and it is small on purpose. Markiewicz has been quoted saying he thinks of Silent Eight as a scale-up rather than a startup, meaning the years of proving the idea are behind them and the years of feeding the machine are what is left. In March 2022 the company closed a $40 million Series B, bringing its total funding to somewhere in the neighborhood of $55 million, and it has spent the money on engineers spread across Warsaw, Singapore, London and New York.

Markiewicz's Twitter handle is @whereiseight, which is the phrasing customers used, back in the early days, when they went looking for the company on Google and could not find it. He has kept the handle. There is something honest about a compliance-software CEO whose social identity is a joke about being hard to see.

A mathematician, three companies deep

He grew up in Poland and studied mathematics there, in a system where public universities are free if you pass the entrance exam. In 2001, still in his early twenties, he started konsultant.it, a small firm that built custom IT solutions for Polish small businesses. He ran it for five years. He then joined Wola Info, at the time one of the larger European IT companies, as strategic sales director, which is the kind of role a certain kind of restless technical founder takes to learn what enterprise procurement actually smells like.

In 2009, he founded SevenFlow Investments, and here the biography takes a turn that surprises people who only know him from the AI-in-banking side. SevenFlow was an engineering business, focused on the technology behind small hydropower plants. Under Markiewicz it grew into the dominant player in the Polish market, expanded into Romania, and eventually became part of a company that reached an IPO. Water, then wires. It is not a straight line, but from a distance the shape is consistent: a technical founder locating a boring, regulated, infrastructure-adjacent market and refusing to leave.

"The smartest people aren't always the most successful. It's the ones who never give up."- Markiewicz, in a 2020 Financial Technology Report interview

The Silent Eight product, plainly

Compliance at a global bank is a factory. Every day, transaction-screening and name-screening systems produce alerts, sometimes millions of them, most of them false. Analysts read the alerts, close the false ones, and escalate the rest. The industry has spent two decades tuning rules to reduce the false positives, and the false positives have not gone away, because the world keeps inventing new names and new payment corridors and new ways for money to look approximately like other money.

Silent Eight puts a model between the alert and the analyst. The model reads the alert the way an analyst would, decides whether the flagged entity is really the sanctioned entity or a namesake, and, critically, writes a narrative explaining its reasoning. The narrative is the product. Without it, the model would be a black box, and a black box cannot pass a regulator's audit. With it, the bank can point at the model's answer and say: here is the entity, here is the reference, here is why we cleared it, here is the timestamp.

Explainability, in Markiewicz's vocabulary, is not a research problem. It is a sales requirement. Banks will not buy AI they cannot explain to their examiner. He has been saying this since before it was fashionable. Watching the rest of the AI industry stumble toward the same conclusion in 2024 and 2025 has been, from Singapore, quietly amusing.

The sibling company

Julia Markiewicz, his sister, is the chief operating officer. She has been there since the beginning. In interviews the two of them tend to describe their working relationship the way people describe a long marriage they still like: there is a division of responsibility, there is trust, and there is a shared understanding that the other person, given the same information, would probably make the same call. Michael Wilkowski, the third co-founder, runs technology. The three-founder shape has held for over a decade, which is unusual.

What he actually talks about

Markiewicz does not do a lot of thought-leadership content. He shows up on the ACFCS member spotlight, on FinTecBuzz, on The Financial Technology Report, and on his own company's podcast, and the through-line of every appearance is the same. He wants banks to have the tools that the consumer internet has been quietly using for years. He notices, correctly, that the recommendation engine deciding what video to show a teenager is more sophisticated than the alert-triage system deciding whether a bank should file a suspicious activity report. He finds this embarrassing on behalf of the industry and treats it as a design opportunity.

He is not, in public appearances, especially performative. He answers questions in complete sentences. He returns to a small number of ideas: explainability, continuous learning, the difference between an alert and a decision, the difference between compliance-as-checklist and compliance-as-model. The interviews are boring in the specific way that engineering interviews are boring. This is a compliment.

"We want to help the 'good guys' win the fight against all the bad actors that are trying to enter the global financial system."- Markiewicz, on the founding thesis

The bet

Regtech in the 2020s has splintered into two categories. There are the vendors selling faster versions of the old rules engine, and there are the vendors selling model-based systems that ask the bank to trust an algorithm's judgment on a regulated decision. The second category is smaller and harder to sell into. Silent Eight is in the second category. Markiewicz's argument, which he has been repeating to procurement committees since 2013, is that the second category is the only one that actually reduces the cost per alert without also reducing the coverage. If he is right, Silent Eight ends the decade as one of the standard tools inside a tier-one bank's compliance stack. If he is wrong, the models never quite pass audit, and the industry stays with its rules engines. He does not appear to entertain the second possibility.

What he does when he is not doing this

He trains. The MMA thing is real. Colleagues have described it as a way of maintaining a cadence that the rest of his life does not otherwise provide. He travels. Silent Eight has offices in four countries, and he moves between them. He reads mathematics for pleasure, which is not surprising, and he learns programming languages the way other people learn spoken ones, which is the joke he sometimes tells about himself. He is 150 employees away from where he started, and he still shows up in podcasts describing the same thesis in the same measured cadence.

The best thing you can say about a founder ten years in is that they have not started saying different things. Markiewicz has not started saying different things.