The Oxford engineer who watched analysts review the same documents tens of thousands of times - then built Arva AI to do it faster, deeper, and with a receipt for every decision.
// Rhim Shah. He holds the same anti-money-laundering credential his software now automates.
Most people who lead a financial-crime product team would never volunteer to sit beside the analysts. Rhim Shah flew to Warsaw and spent hundreds of hours doing exactly that. He watched people open a document, check a name, cross-reference an ownership chart, flag a risk, close the file, and start again. Tens of thousands of times a day across the team. The work was, in his own description, mundane and repetitive. That observation is the seed of Arva AI.
Today Shah is co-founder and CEO of Arva AI, a company that builds auditable AI agents to handle the back office of fighting financial crime - the AML reviews, the Know Your Business checks, the transaction monitoring, the adverse-media sweeps. The pitch is not that the AI is clever. The pitch is that the AI is tireless, contextual, and leaves a trail a regulator can follow. Arva says its agents automate somewhere between 80% and 92% of the manual review work that used to swallow whole teams.
"Fighting financial crime shouldn't be reactive or slow. It should be intelligent, contextual, and auditable."
The credential matters here. Shah is a Certified Anti-Money Laundering Specialist. He is not an outsider promising to disrupt a field he has never touched - he holds the exact certification the people doing the work hold. That is the difference between a founder who has read about compliance and one who has been audited by it. When he talks about guardrails and auditability, he is describing the thing that would have made his old job survivable.
Shah studied engineering at the University of Oxford. He did an associate product manager internship at Google, then led multiple product teams at Jobandtalent, then ran the FinCrime product team at Revolut Business. The Revolut chapter is the one that mattered. He has said he chose its startup chaos over Google's corporate structure on purpose - weekly access to the CEO, real ownership, faster learning. The decision to trade comfort for velocity is a pattern with him, not a one-off.
Figures as described by Arva AI across KYB, transaction monitoring and screening workflows.
The Warsaw shadowing was not Shah's first business instinct. While still at Oxford he launched Carbon Codes, a sustainable food marketplace for students, and later sold it to a competitor. Earlier still, he had self-published study guides for university entrance exams. The anecdote he tells about those guides is small and revealing: offered roughly £50 for the rights, he turned them down, published them himself, and - by his telling - earned 100 to 1,000 times more. He was, in other words, already someone who could spot mispriced work and refuse the easy exit.
That instinct shows up again in how he frames Arva's ambition. He is not chasing a quick flip. He talks about a profitable, durable business on a five-to-ten-year horizon, aimed at what he calls a $10 to $100 billion problem buried in the back office of every financial institution. Compliance, in his framing, is not a cost center to be tolerated. It is a bottleneck to be turned into a growth engine.
"It's a $10 to $100 billion problem."
Shah did not build Arva alone. His co-founder and CTO is Oli Wales, a Cambridge computer-science graduate who was lead product engineer at Opvia and a full-stack engineer at The Trade Desk. Oxford and Cambridge, product and engineering, the classic complementary split. The two went through Y Combinator's Summer 2024 batch and raised a seed round - about $3M, roughly $3.13M in total funding - led by Gradient Ventures, Google's AI fund. The company already counts leading banks and fintechs in the UK and US among its customers, and it celebrated early by putting Arva on a New York Times Square billboard.
It is easy to say AI will replace compliance analysts. It is much harder to do it in a domain where a wrong call is not an inconvenience but a regulatory breach. That is why Shah's language keeps circling back to the same three words: intelligent, contextual, auditable. An agent that automates a review but cannot explain itself is useless to a bank. The value is not just the speed. It is the audit trail that travels with every decision, so a human and a regulator can both see how the machine reached its conclusion. Shah's advice to other founders in the space reflects this discipline - understand the compliance nuances deeply, stay close to customers, build for scale, and put the guardrails in first.
For someone who spent hundreds of hours watching the most repetitive job in banking, the goal was never to make analysts disappear. It was to give them back the hours they were spending on the mundane, and to hand the dull, careful, endless cross-checking to something that does not get bored. That is the quiet ambition under the funding headlines: not the end of compliance work, but the end of compliance drudgery.
Arva frames itself as an AI workforce rather than a single tool, and the distinction is deliberate. A bank does not have one compliance problem. It has a stack of them, each with its own rules and its own paper trail. There is Know Your Business, the work of verifying that a company is what it claims to be, untangling ownership structures and enriching entity data. There is Know Your Customer and the onboarding flow behind it. There is transaction monitoring, the endless watch for the one payment among millions that looks wrong. There is adverse-media screening, sanctions and PEP checks, and the ongoing duty to keep watching a customer long after they signed up. Arva's platform is built to let a bank deploy, monitor and audit agents across all of it, which is why Shah talks about building, deploying and monitoring an AI workforce rather than shipping a feature.
The headline number Arva uses - automating as much as 92% of reviews and cutting review times by a similar margin - is less interesting than the second half of the sentence. The company says it strengthens controls while it speeds them up. In most software, speed and rigor pull against each other. In compliance, a tool that is fast but sloppy is worse than no tool at all, because it manufactures false confidence at scale. Shah's wager is that a well-built agent can be both quicker and more thorough than a tired human on hour seven of document review, precisely because it does not tire.
Arva sits in a corner of fintech that rarely gets the glossy coverage - regtech, the unglamorous business of regulatory technology. But the timing is hard to argue with. Generative AI got good enough to read messy documents and reason about them at roughly the same moment that banks were drowning in compliance headcount costs and regulators were tightening expectations around auditability and model risk governance. Shah, with his machine-learning background and his frontline view of the bottleneck, was standing exactly where those two trends met. The insight was not that AI was powerful. It was that AI had finally become reliable enough to be trusted with work where being wrong is expensive.
That is also why the auditability obsession is not a marketing line. Banks cannot adopt a black box. Every decision an Arva agent makes has to be explainable and traceable, because somewhere down the line a regulator may ask the bank to justify it. The company's whole proposition rests on the idea that AI judgment and human judgment are not rivals but layers - the machine handling the volume, the human handling the edge cases, and a clean record connecting the two. For a founder who once held the CAMS certification and lived inside those workflows, designing for the auditor is not a constraint bolted on at the end. It is the starting point.
Put the pieces together and a pattern emerges. A teenager who refused a £50 buyout and published his own work. A student who built and sold a company before graduating. An operator who picked the harder, faster job twice. A leader who flew to Warsaw to watch the work nobody wanted to watch. None of it was obvious at the time, and none of it looked like a straight line. But it added up to a person uniquely placed to look at the most tedious corner of banking and see, instead of drudgery, a $10 to $100 billion company waiting to be built. Arva AI is young, and the work of fighting financial crime is older than any of its founders. The bet is that the next chapter of that work belongs to agents that read, reason, and never get bored - with Rhim Shah writing the operating manual.
Chose Revolut's chaos over Google's ladder. Trades safety for learning speed, on purpose.
Hundreds of hours shadowing analysts in Warsaw before writing a line of the pitch.
Refused the £50 buyout. Built for himself. The instinct that spots mispriced work.
In a regulated field, auditability is the feature. Build the safety, then the speed.
Action and rapid iteration beat polish. Ship, learn, repeat.
Create value for customers, and the team grows from it.
Open communication and visibility across every level.