At OpenSanctions, we believe that replacing proprietary datasets and black-box methodologies with an open-source approach is a logical step for the compliance industry. It provides full auditability and explainability, ensuring that any datapoint can be traced back to its source.
That ethos extends to our entire tech stack: our screening API service, yente, is also fully open-source. This means that not only can you download the software and run it within your own infrastructure, but you can also modify and extend it to perfectly fit your use case.
Fellow open-source enthusiasts at Marble have done precisely this. Marble, based in Paris, is an AI decision hub that connects all Fraud and Anti-Money Laundering (FRAML) data sources under one open, self-improving architecture. As their customer base grew, so did the demands of real-time transaction screening — simultaneously checking thousands of identities against sanctions, PEPs, and watchlists.
In what began as a personal project, Marble engineer Antoine Popineau set out to push yente's matching performance further. Rather than designing a new scoring methodology from scratch, he built directly on yente's core matching and scoring algorithms — extending an open foundation rather than replacing it.
The result was Motiva: a reimplementation of yente's matching algorithms, purpose-built for high-concurrency environments. Marble has adopted and now sponsors its continuous development.
Rebuilding the engine, not the logic
yente is built in Python, with nomenklatura as its scoring engine. It provides functions to search for, retrieve, or match FollowTheMoney entities, including people, companies, or vessels that are subject to international sanctions. We roll out updates regularly to add new features, expand our data model, and improve the scoring system.
As Marble’s customer base expanded and request volumes grew, it needed an upgrade that would consistently perform under high-concurrency entity screening and provide future scalability. Rather than build new logic, Marble took OpenSanctions' core matching and scoring algorithms and rebuilt the engine using the fast, memory-efficient programming language Rust.
"I didn't want to reinvent the wheel, since OpenSanctions had already done a great job at providing battle-tested matching algorithms. Also, staying faithful to the underlying logic gave me a way to verify that Motiva was correct – so staying faithful was both principled and efficient. Exactly what we at Marble wanted to achieve: pushing yente further."
Antoine Popineau, Backend Developer, Marble
Motiva reimplements all three of OpenSanctions' matching families — NameBased, NameQualified, and logic-v1 — and reuses our rigour-based text normalization functions for cleaning and standardizing names.
Marble could also check their work against the original rather than trusting it blindly. To maintain the same matching quality and precision as OpenSanctions' reference implementation, integration tests compare Motiva's scores directly against nomenklatura and run automatically on every code change to ensure compatibility.
Marble goes into more detail on the technical details here.
A natural upgrade path
Marble has seen real, measurable impact from the rollout of Motiva. Memory consumption now stays stable regardless of concurrent request count, and P99 latencies remain consistent during traffic spikes. Resource utilization is also more efficient, meaning Marble can handle higher throughput on fewer instances. The operational cost savings from this will compound as transaction volumes grow.
“Open infrastructure lowers the cost of building compliance infrastructure,” says Friedrich Lindenberg, Founder of OpenSanctions. “Marble didn't need to design or license a new matching methodology, or reverse-engineer a black box — they could rebuild yente’s matching layer to suit their specific use case.”
“This kind of collaboration creates a natural upgrade path that strengthens the broader compliance ecosystem. It encourages open knowledge sharing between organizations, and ultimately allows us to work together to fight financial crime.”
Friedrich Lindenberg, Founder of OpenSanctions
The relationship isn't one-directional: we’ve continued to advance our own reference implementation in both speed and matching quality – reflected in the updates we’ve rolled out over the last year. This open exchange has enabled us to learn from Marble’s work on Motiva, too – we’ve since used Rust in the most performance-critical components of rigour and nomenklatura.
Our logic-v2 matching algorithm was designed to be more precise, explainable, and culturally aware, reducing false positives and improving cross-language matching. And, thanks to the name-matching part of OpenSanctions’ rigour version now being written in Rust, it will be easily adapted into Motiva, too.
Meanwhile, a new performance-driven version of yente (v5.4.0) was released earlier this year, and, with the upgraded rigour and nomenklatura stacks, end-to-end matching throughput on the /match endpoint became roughly ~50% faster.
The power of open infrastructure
Marble’s reimplementation of yente shows that open data alone wouldn't have made this possible; the infrastructure that data sits within also needs to be open. Marble engineers could deploy, extend, and validate against our methodology without needing direct support. It's a minor detail with a major impact — open source enables technological advancement only when it's genuinely easy to integrate and adapt.
yente comes with ready-to-run, pre-built Docker containers and is released under an open-source license — removing the usual barriers that limit technologists in the industry. While we built it for OpenSanctions data, it isn’t limited to it. It’s used to screen proprietary datasets, too.
In a traditionally siloed industry, open infrastructure creates a range of opportunities for individual engineering teams. But the pay-off is even bigger than that: fighting financial crime is fundamentally collaborative. Open knowledge sharing and collaboration between organizations means we’re constantly raising the bar and improving compliance technology.
Thanks to Marble’s reimplementation, any team looking to build on top of yente now has a validated blueprint to follow.
What will you build?
Antoine Popineau has open-sourced Motiva here. Read Marble's case study here.
