The first product Baran Ozkan wanted to build came from an hour lost at the Louvre. He had waited roughly thirty minutes for an audio-guide headset, discovered it did not work, then queued again for a replacement. The irritation followed him out of the museum. Why, he wondered, could a phone not do the same job?
His answer was Audiotap, a mobile travel companion designed to replace bulky museum hardware. It did not survive. The text-to-speech technology sounded stiff. The business model had holes. Ozkan, an engineer who loved making things, had not yet learned how to raise capital, sell a product or find the customers who might sustain it. He shut the project down.
The failed app matters because it gives the polished Flagright story a useful scratch. Ozkan did not glide from engineering into a Y Combinator company. He first built a solution before he understood the buyer, the timing or the economics. Years later, when he started again, he reversed the sequence: live inside the problem, learn its constraints, then build.
“Action beats information.”Baran Ozkan on validating an idea
An engineer enters the regulated world
Ozkan is originally from Turkey and trained as an engineer. He moved to the United States for a master's degree, then spent about five years working as a software engineer. His first role was in healthcare insurance, where reliability and accountability were operating conditions rather than values printed on a wall. He later moved through fashion, e-commerce, marketing and logistics. The industries changed. The habit of looking for the gap between a process and its software stayed.
That path eventually took him to Germany and Lithuania, and then into product leadership at the money-transfer company TransferGo. As director of product, he worked around identity verification, fraud prevention and money-laundering controls. His job put him on the buying side of financial-compliance software, close enough to see the difference between a confident sales demonstration and a system that had to make decisions under production pressure.
For roughly fifteen months, he searched for a real-time, risk-based transaction-monitoring product. The field split into familiar frustrations: tools priced for large banks, interfaces that burdened investigators, and products with weak APIs or developer documentation. Some appeared capable in a demo and disappointed in testing. He did not need to invent a customer persona. He was the frustrated customer.
Ozkan's second company began with procurement notes, not a blank whiteboard. Failed vendor evaluations became product requirements: real-time decisions, clear APIs, usable investigations and pricing a smaller financial company could absorb.
The friend who could build the other half
Ozkan took the problem to Madhu G. Nadig, a former teammate and close friend with experience building data-intensive and real-time systems. They founded Flagright in 2021. The pairing was legible: Ozkan knew the workflows, the buyers and the commercial pain; Nadig could lead the engineering and design of the infrastructure beneath them.
Their friendship was not incidental to the org chart. Ozkan has argued that co-founders should be friends because a startup guarantees hard conversations. Colleagues can disagree. Friends have a path back after the disagreement. It is a personal theory of institutional resilience: the relationship needs enough history to hold when the company supplies heat.
The first version came quickly. The company began serving customers early, joined Y Combinator's Winter 2022 batch and reported eight startup customers by June of that year. Within six months, Ozkan later recalled, it had traction on three continents. The product expanded from transaction monitoring into risk scoring, watchlist screening, case management, investigations and reporting.
The closed loop Ozkan wants compliance software to run
The product is the explanation
Ozkan often compresses his view of finance into four words: “Trust is the product.” The phrase sounds abstract until it is translated into software. A payment system has to decide quickly enough to avoid freezing ordinary life. A compliance team has to understand why an alert fired. A manager has to know who changed a rule. An auditor has to reconstruct the decision later. Speed without a record is fragile; a perfect record produced too late is not much use either.
This is also why his AI argument is narrower than the marketing language surrounding the technology. He does not frame the machine as a substitute for every analyst. Flagright uses rules for known risks and AI for contextual work around investigations, evidence and prioritization. Humans keep control where risk is highest. Proposed changes can be replayed against historical alerts, compared with earlier decisions and versioned before release.
The distinction becomes clearer in an example. A retired customer who rarely withdraws cash suddenly uses an ATM at midnight for a large amount. A static rule can spot the amount and hour. A useful system also knows the customer's normal behavior, assembles the relevant context, routes the case and records what the institution did. The alert is merely the doorbell.
“Compliance requires explainability and a certain amount of mechanical oversight.”Ozkan on the boundary between AI and judgment
Twenty edits, one fixed problem
Early certainty did not extend to the business model. Ozkan has said Flagright changed its monetization approach around twenty times in its first year. Prospects found one proposal confusing, suggested another structure, then reacted again. Eventually the company settled into annual contracts. The willingness to revise pricing sat beside a stubbornness about the underlying pain.
This is his most transferable founder distinction: conviction belongs to the problem, not the first packaging of the answer. He recommends showing prospective customers something small, even broken, because a real artifact produces information. Friends and family can offer encouragement. A buyer willing to sign a letter of intent, test a workflow or pay supplies evidence.
The same experimental logic shaped spending. Ozkan has described testing acquisition channels, watching for return, then putting more money behind what worked. He personally closed the company's first $2 million in annual recurring revenue before building out a sales team. “Depth is underrated,” he tells would-be founders, a compact objection to the tax of chasing every new trend.
A reported customer curve
The CEO in the support queue
There is an unusually revealing detail in Ozkan's account of Flagright's growth. His title in the company's Slack is “Customer support,” not CEO. Clients can tag him without first decoding the hierarchy. Internally, customer channels sit in a group with notifications turned on. Responsiveness is meant to be everyone's reflex.
For complex enterprise software, this is product strategy disguised as manners. Financial institutions do not finish buying at signature. They integrate data, tune rules, review alerts and answer audits. A fast response at the tense moment becomes part of what they purchased. Ozkan says word of mouth is Flagright's strongest sales channel, and the mechanics are easy to understand: people change employers, remember which vendor answered and carry that trust into the next buying room.
By 2025, Flagright said it served more than 50 customers across six continents. In April that year, it announced a $4.3 million seed round led by Frontline Ventures. By June 2026, the company reported more than 100 financial institutions across over 30 countries and raised a $12.5 million Series A led by Infinity Ventures, with Sella and continuing support from Frontline and Y Combinator.
The round is intended to expand explainable AI across investigations, alert intelligence, rule optimization and decision support, while increasing Flagright's presence in the United States. That plan moves the company toward larger institutions and a harder version of the same founding test: can a risk team explain what is happening more clearly after the software arrives?
From clearing queues to asking better questions
Ozkan's ambition is to make Flagright a default choice for financial institutions that take financial-crime prevention seriously. More interesting is the operational picture beneath that goal. He wants teams to spend less of the day clearing noisy queues and more of it asking what their business, customers and exposure are telling them. The aspiration is proactive risk management: bad actors have fewer places to hide while legitimate customers encounter less friction.
That future still contains people. Analysts oversee consequential decisions. Regulators can follow the trail. Institutions choose their risk appetite. AI absorbs repeatable labor and organizes context, but it does not erase responsibility. The product Ozkan could not buy was real-time monitoring. The company he is now building has a larger brief: make speed and accountability coexist.
The distance from a broken Louvre headset to a compliance platform is wide, but the two projects share a founder's itch. Something slow, awkward or unreliable should work better. The difference is everything Ozkan learned between them. With Audiotap, irritation supplied the whole thesis. With Flagright, irritation met years of domain experience, a buyer's scar tissue and a friend who could build the systems underneath it. The second time, the complaint came with a map.
Links