How Wise Is Building a Scalable 2025 Tech Stack with Kubernetes and AI

Wise Engineering: 2025 Tech Stack Update

By the 2024 fiscal year, Wise serves 12.8 million active customers, processing up to £30 billion in cross‑border transfers each quarter. Over 60% of transfers settle instantly, powered by a technology‑first philosophy, robust architecture, and a dedicated engineering organization.

Engineering Organization

Wise employs more than 850 engineers worldwide, organized into autonomous squads and tribes that are empowered to innovate and make independent decisions, fostering transparency, trust, and collaboration.

Web and Mobile Applications

The web platform runs on CRAB, a Wise‑specific abstraction built on Next.js, comprising 40 independent apps that isolate product functionality for safer, more manageable deployments.

Testing now heavily uses Storybook together with Chromatic to capture visual snapshots of React components, preventing regressions before code reaches customers.

Mobile – iOS : iOS engineers migrated 250+ Xcode modules from Xcodegen to Tuist and switched from CocoaPods to Swift Package Manager, cutting build time from 28 seconds to 2 seconds and moving toward a composable Swift architecture.

Mobile – Android : Android engineers maintain a codebase of 300+ Gradle modules (~1 M lines of code) covering two production apps, six sample apps, 17 JVM modules, 221 Android modules, and 65 multi‑platform modules. Key improvements include shared BFFs across platforms, KSP‑based code generation, and exploration of Kotlin Multiplatform. The UI has fully transitioned to Jetpack Compose, using Kotlin 2.x, coroutines, and MVVM patterns.

Backend Services

Wise runs over 1,000 backend services primarily written in Java and Kotlin. Recent focus has been on internal tooling for automation, scalable dependency management, and a minimal‑configuration microservice framework that standardizes security, observability, database access, and Kafka integration.

Introduced an internal microservice framework released as a component to accelerate standard service creation.

Standardized build pipelines with internal Gradle plugins, simplifying GitHub Actions workflow updates across 700+ Java repositories.

Built a language‑agnostic automation service for large‑scale codebase changes and centralized Java dependency upgrades.

Payments Integration

Wise has launched instant payment systems such as InstaPay in the Philippines and gained access to Japan’s Zengin and Brazil’s PIX networks.

Network connectivity is centralized via AWS Transit Gateways, with the Australian data center being one of the first AWS Outpost Server deployments, ensuring consistent tooling across regions.

Public API

The public REST API, secured with OAuth, enables enterprises to integrate cross‑border payments, currency conversion, and account management, backed by comprehensive documentation and developer tools.

Infrastructure Platform Expansion

To support rapid growth, Wise rebuilt its infrastructure on a Kubernetes‑based Compute Runtime Platform (CRP), replacing the previous Terraform/JSONNET/Concourse stack.

Terraform remains for IaC, but the codebase was rewritten for greater flexibility.

RKE2 handles cluster bootstrapping; Rancher manages overall cluster state.

Helm replaces JSONNET for better maintainability and upstream compatibility.

ArgoCD with custom plugins ensures fully automated configuration and consistency.

Envoy service mesh now provides seamless service discovery and integration.

Kubernetes clusters grew from 6 to over 20 while maintaining efficiency.

CRP also introduces smarter autoscaling and cost‑optimization, including a flexible autoscaler, vertical pod autoscaling (Goldilocks), and KEDA‑driven horizontal scaling.

Data Infrastructure

Wise’s data backbone relies on MariaDB, Postgres (migrating to Amazon RDS), MongoDB Atlas, and Redis, with ongoing exploration of distributed databases.

Kafka powers real‑time data movement, with increased cluster capacity and rack‑aware replicas for fault tolerance.

Internal data‑movement services stream data to Snowflake, S3 Parquet, Iceberg, etc.

Over 1 trillion records are archived across databases, reducing storage costs and simplifying backup.

Analytics uses Snowflake, an S3‑based data lake with Apache Iceberg, and Trino for unified querying across Iceberg, Snowflake, and Kafka streams. Workflows are orchestrated with Airflow and dbt‑core, while dashboards are built in Looker or Superset.

Artificial Intelligence

Wise’s ML stack supports both exploration and production, using Amazon SageMaker Studio for experiments, EMR + Spark for large‑scale processing, and Airflow for pipeline orchestration. Feature stores, MLflow, and Ray Serve enable efficient model training, tracking, and deployment.

A secure gateway connects to LLM providers (Anthropic, AWS Bedrock, Google Gemini, OpenAI) via a LangChain‑inspired Python library, and a custom RAG service enriches responses with internal documents.

CI/CD Evolution

Migration from CircleCI to GitHub Actions, combined with detailed metrics, reduced build times by 15% (≈1 000 hours saved monthly on 500 k builds). The SLSA framework is being incrementally adopted for supply‑chain security.

Deployments shifted from Octopus to Spinnaker, enabling automated canary analysis (5% traffic routing, 30‑minute metric evaluation, automatic rollback), preventing hundreds of risky releases in 2024.

Observability Stack

Wise moved from Thanos to the Grafana LGTM stack (Loki, Grafana, Tempo, Mimir), processing ~6 million metric samples per second and 150 million active series for the largest tenant. Additional pilots with Grafana Pyroscope explore performance insights.

Dedicated observability CRP clusters provide out‑of‑the‑box monitoring for services across environments, simplifying setup and reducing manual effort.

Conclusion

The 2025 tech stack demonstrates Wise’s commitment to delivering the fastest, most reliable, and cost‑effective cross‑border money transfers for 12.8 million active customers, with a focus on standardization, scalability, risk management, and continuous innovation across frontend, mobile, backend, data, AI, and infrastructure.