AI-Driven Risk Intelligence: Graph, Shadow Testing, and MLOps

The client is the global digital-payments two-sided network: hundreds of millions of active accounts across roughly two hundred markets, with trillions in total payment volume and tens of billions of payment transactions annually. The global payments company owns their payment processing platform, the P2P platform, their international money transfer service, the guest checkout product, their digital currency, and a BNPL business at tens of billions in 2024 BNPL volume. Taller's engagement at this organization is the largest in the portfolio. Six master-deck case studies cover the engagement across the P2P platform, the global payments company, and their payment processing platform deliverables. The four cases below address the global payments company-specific case studies the dossier proposed.

The global payments company's risk stack — the custom real-time graph platform with Gremlin query interface, graph embeddings on billions of vertices, their shadow-testing environment for regulated fraud models, their MLOps platform fabric since 2022 — is architecturally distinctive. The risk function requires ongoing AI/ML engineering capacity for model development and ML platform engineering at production scale.

Taller engineers contribute capacity to the global payments company's Agent Builder — the self-service agent platform inside the global payments company — and to the Machine Learning Engineer surface supporting large-scale ML infrastructure for online recommendation, ads ranking, personalization, search, and NLP on Spark, EMR, and Kafka. The capability reference Taller carries on production anomaly detection at scale is the cybersecurity company engagement: a backend and data-science team that developed anomaly-detection models capturing insights from endpoint snapshots, on Python and ML models against multi-source telemetry.

Same-client AI/ML engineering contributor presence into the global payments company's risk and ML teams. The cybersecurity company anomaly-detection engagement delivered enhanced threat-detection accuracy, reduced response time to threats, and increased security coverage across endpoints.

The global payments company's risk outcomes — a seven percent performance gain in a collusion-fraud model via graph embeddings; transaction loss rate held flat at 0.07 percent of TPV while TPV grew seven percent in 2025; combined transaction-and-credit-loss rate at 0.10 percent of TPV — are the global payments company risk-team outcomes earned by the graph platform with Gremlin, their shadow-testing environment, and their MLOps platform architectures that this company built and operates. The architecture is the global payments company's. Taller's contribution is contributor capacity into the AI/ML engineering supporting the risk and ML functions, plus the cybersecurity company anomaly-detection reference for the capability shape.