01Who they are right now
Somewhere on the fifth floor of a federal building this morning, an analyst opens a case. There are 32 spreadsheets, three transaction databases, a few KYC PDFs, a stack of suspicious-activity reports, and an inbox full of tips. The deadline was yesterday. They click into DataWalk. Within minutes, the noise becomes a map - people, accounts, addresses, shell companies, and the lines connecting them. The case stops being a pile and starts being a question they can finally answer.
DataWalk is the software in that scene. A graph- and AI-enabled investigation platform built so the people whose job is to find fraud, money laundering, and bad actors can stop fighting their own tools. Headquartered in Redwood City, with engineering and origins in Wrocław, Poland - a Polish-American company that managed to stay obscure while landing inside the U.S. Department of Justice, the Department of Homeland Security, the Department of State, USDA, and a small list of large global banks.
The pitch in one sentence: stop investigating the data. Start investigating the case.
02The problem they saw
Investigators do not have a data shortage. They have a translation shortage. The fraud team has the transactions. KYC has the identity documents. The watchlist team has the sanctions hits. Compliance has the alerts. None of those systems were built to speak to each other, which is why analysts spend the bulk of their hours moving data between tabs and not, you know, catching anyone.
This is the problem DataWalk's founders kept running into in the early 2010s in Wrocław. They had a hunch that the answer was not another dashboard. The answer was a graph - a model that treats every person, account, transaction, document, and event as a node, and every relationship between them as an edge. Connect them once, query them visually, and a case stops being a research project. It becomes a path.
Easier said than built. Graph databases existed. Link analysis existed. What did not exist - at least not in a way an actual fraud analyst could use without three engineers babysitting - was a system that handled the messy reality of enterprise data: hundreds of millions of records, dirty fields, mismatched identifiers, dynamic feeds, security constraints, regulatory reporting, and the patience of an investigator who would very much like an answer before lunch.
03The founders' bet
The company began life in 2011 as PiLab, founded by Paweł Wieczyński, Krystian Piećko, and Sergiusz Borysławski. Wieczyński, an economist by training with a background in logistics and quality management, ran the business. Piećko, the CTO, was the one who built the core. In 2016, MIT Technology Review named him one of its "Innovators Under 35" for his work on data representation. The same year, the founders opened DataWalk Inc. in Silicon Valley, hired veteran tech executive Gabe Gotthard - a man who once helped sell 3PARdata to HP for $2.5 billion - and pointed the product squarely at the U.S. market.
This is the unusual bet at the heart of the company: keep engineering in Poland, where the talent is dense and patient, while running commercial operations in California, where the customers are. Most startups split this way collapse under the weight of the time zones. DataWalk did not. The product survived its translation from a Polish lab to American agencies.
A 15-year footnote, in chronological order
04The product
What an analyst actually does with DataWalk:
They point it at every source they have - transaction systems, watchlists, KYC records, blockchain feeds, external intelligence, CSVs that someone in another department called "the database" - and let the platform unify them. Entity resolution runs at scale, collapsing duplicate identities. A knowledge graph forms. From there, the analyst runs queries by clicking, not by writing code: find every account connected to this person within three hops, score the risk, flag the anomalies, save the path.
It is no-code where no-code matters, and code-friendly where it has to be. There is an investigation sandbox for hypothesis testing, AI explainability so investigators can defend their findings to a regulator or a judge, multi-hop tracing for cryptocurrency cases, and the boring-but-essential machinery of audit logs, role-based security, and regulatory reporting. It runs in the cloud, or on-premise behind whatever firewall a three-letter agency requires.
The trick is not any one feature. The trick is that it was designed by people who took the unglamorous parts of an investigator's day seriously.
05The proof
The customer list is the argument. ING Bank. The U.S. Department of Justice. Department of Homeland Security. Department of State. USDA. Insurers and intelligence agencies that, for understandable reasons, do not put their names in press releases. These are not customers who pick a tool because the slide deck was nice. They pick a tool because their analysts stopped complaining.
The lean-team trick
06The mission
DataWalk's stated mission is short: turn complex data into adaptive contextual intelligence so humans and AI can act with clarity and speed. The honest version is shorter: give the people chasing bad actors a chance to keep up. AML alerts rose. Fraud sophistication rose. Crypto added a new dimension. Sanctions regimes complicated overnight. Investigators were asked to find more, faster, with the same headcount. Someone had to build the software that made that possible without doubling the engineering staff at every customer.
The company's other quiet mission - though it rarely says so out loud - is to prove that a serious enterprise platform can come from somewhere other than Silicon Valley. Wrocław as a graph-analytics capital is not a phrase you have heard before. Give it time.
07Why it matters tomorrow
Two trends decide DataWalk's next decade. The first is AI. Generative models and agentic systems are perfectly suited to investigation work - summarize a case, suggest the next hop, draft the regulator-facing narrative - but only if they are anchored to a clean knowledge graph that they cannot hallucinate around. DataWalk happens to have spent fifteen years building exactly that anchor. The second is regulation. Every year, more of the financial system gets pulled into AML, KYC, sanctions, and crypto-tracing obligations. Every year, the cost of getting any of it wrong goes up. Tools built around the investigator, not around the database, become structural rather than optional.
Back to the analyst on the fifth floor. The case that was a pile is a graph. The deadline that was yesterday is a meeting at four. The three engineers who used to be on call to write SQL are working on something else. The investigator clicks. The connections render. The path to the suspect, the one they would not have found scrolling through a spreadsheet, is visible. They send the report. They go home on time.
That is the product. The rest is footnotes.