Yassine LASRI

Product Manager at @SynaptiqueM, IT & Telecoms Professional, Interested in AI, IoT, 5G, Big Data, RA, FMS, NG B/OSS, Mobile Money

[Yassine LASRI @ylasri]

@MuhiziSa @ARCTBDI @EacoInfo Good job Mr Samuel.

[Yassine LASRI @ylasri]

@ImaneOuaadil @aminedinar Bonne nouvelle alors, je viens de payer 1500 MAD pour un visa express avec beaucoup de paperasse. Ca serai bien de supprimer ce visa à l'instart du Togo et Bénin

[Gautam Kedia @thegautam]

TL;DR: We built a transformer-based payments foundation model. It works.

For years, Stripe has been using machine learning models trained on discrete features (BIN, zip, payment method, etc.) to improve our products for users. And these feature-by-feature efforts have worked well: +15% conversion, -30% fraud.

But these models have limitations. We have to select (and therefore constrain) the features considered by the model. And each model requires task-specific training: for authorization, for fraud, for disputes, and so on.
Given the learning power of generalized transformer architectures, we wondered whether an LLM-style approach could work here. It wasn’t obvious that it would—payments is like language in some ways (structural patterns similar to syntax and semantics, temporally sequential) and extremely unlike language in others (fewer distinct ‘tokens’, contextual sparsity, fewer organizing principles akin to grammatical rules).

So we built a payments foundation model—a self-supervised network that learns dense, general-purpose vectors for every transaction, much like a language model embeds words. Trained on tens of billions of transactions, it distills each charge’s key signals into a single, versatile embedding.

You can think of the result as a vast distribution of payments in a high-dimensional vector space. The location of each embedding captures rich data, including how different elements relate to each other. Payments that share similarities naturally cluster together: transactions from the same card issuer are positioned closer together, those from the same bank even closer, and those sharing the same email address are nearly identical.

These rich embeddings make it significantly easier to spot nuanced, adversarial patterns of transactions; and to build more accurate classifiers based on both the features of an individual payment and its relationship to other payments in the sequence.

Take card-testing. Over the past couple of years traditional ML approaches (engineering new features, labeling emerging attack patterns, rapidly retraining our models) have reduced card testing for users on Stripe by 80%. But the most sophisticated card testers hide novel attack patterns in the volumes of the largest companies, so they’re hard to spot with these methods.

We built a classifier that ingests sequences of embeddings from the foundation model, and predicts if the traffic slice is under an attack. It leverages transformer architecture to detect subtle patterns across transaction sequences. And it does this all in real time so we can block attacks before they hit businesses.

This approach improved our detection rate for card-testing attacks on large users from 59% to 97% overnight.

This has an instant impact for our large users. But the real power of the foundation model is that these same embeddings can be applied across other tasks, like disputes or authorizations.

Perhaps even more fundamentally, it suggests that payments have semantic meaning. Just like words in a sentence, transactions possess complex sequential dependencies and latent feature interactions that simply can’t be captured by manual feature engineering.

Turns out attention was all payments needed!

[Yassine LASRI @ylasri]

@SelemaniMarina Raconte les détails :)

[Yassine LASRI @ylasri]

@SelemaniMarina 2 RDV dans la même journée à Kinshasa c'est déjà une bonne performance

[Yassine LASRI @ylasri]

@benjimk Jolie

[Yassine LASRI @ylasri]

@jisaacmercier Non. Loin de ça, juste longue distance pour faire le tour de Kin a Goma ...

[Yassine LASRI @ylasri]

@Bibimusafi Bien dommage que c'est à plus de 2000km de Kinshasa, c'est bien cela ?

[Yassine LASRI @ylasri]

@DiavitaMyriam @SuminwaJudith @Chrisbalembo @AlTshibanda @Raissabulela1 Good job engineer

[Yassine LASRI @ylasri]

@PriscaEfomi24 Beaucoup de courage à vous. Il faut avancer

[Yassine LASRI @ylasri]

@tisyamukuna Est ce qu'il y a des offres de vente de cette variété sur Kinshasa ? Quel est le prix par kg ?

[Yassine LASRI @ylasri]

@Raissabulela1 100%😎

[Yassine LASRI @ylasri]

@LeonJuste @CongoOfficiel @MTN_123 @airtel_congob @PR_Congo @GouvernementCG @PostesTelecomcg Félicitations aux congolais et aux congolaises

[Yassine LASRI @ylasri]

@Raissabulela1 @SuminwaJudith Statistiquement vous avez plus de chances d'avoir une répondre à partir de la 7eme relance

[Yassine LASRI @ylasri]

@SuminwaJudith @Forbes Félicitations madame la Première Ministre.

[Yassine LASRI @ylasri]

@KasekaSandrine C'est où exactement svp ?

[Yassine LASRI @ylasri]

Après un voyage de 2 semaines avec @royalairmaroc retour sur Casablanca, surprise baggage non encore localisé. Bravo #RAM pour cette médiocrité

[Yassine LASRI @ylasri]

@ColucciLaetitia C'est bien ça Kin. Enjoy it

[Yassine LASRI @ylasri]

Voici la réponse de dame @royalairmaroc : Nous reviendrons vers vous dès qu'il y aura une information de l'aéroport.
Bien à vous !
Depuis sans aucune nouvelle !

[Yassine LASRI @ylasri]

1400$ pour un A/R et 2 bagages d'une valeur de 1500$ perdus par @royalairmaroc! Ils me disent que mes bagages arriveront à destination initiale après mon retour, soit une semaine après. Déception totale, service client inexistant. Voyageurs, soyez avertis ! #Scandale #ServiceClient

[Yassine LASRI @ylasri]

@RAM_Maroc_CEO @adelfakir