Airwallex is building the future of global banking, helping businesses move money across borders, manage multiple currencies and operate internationally. As the company scaled, they needed a data platform that could keep pace with real-time financial operations, whether it was powering trusted dashboards, self-service analytics or personalized customer experiences. But the use of open source software (OSS) tooling was resource intensive to scale, pulling engineering time into managing infrastructure as opposed to higher valued tasks. By consolidating their data and AI stack on Databricks and Google Cloud, Airwallex now operates from a single source of truth while reducing source-to-gold latency to a mere 15 minutes.
Enabling analytics, personalization and customer self-service
Founded in 2015 in Melbourne, Australia and now co-headquartered in Singapore and San Francisco, Airwallex helps businesses manage their global banking and financial operations. As global commerce accelerates and finance teams expect AI-driven automation, the fintech company wanted to embed intelligence into their platform to enable faster, leaner operations. To keep pace with these aspirations for growth, Airwallex’s data team needed to scale a mix of internal and customer-facing initiatives that relied on governed, high-confidence data. Internally, the team was focused on automating operational processes and scaling business-critical dashboards that leadership and teams could trust for decision making. This included management dashboards for tracking company performance and operational dashboards for monitoring product rollouts and day-to-day platform health, such as measuring adoption and performance for new capabilities like their AI assistant.
Externally, Airwallex wanted to expand customer-facing self-service experiences and analytics. This way, customers could access insights and resolve common needs without relying on the customer support team. Delivering on that vision required reliable, scalable data serving, along with stronger trust signals across customer communications to improve the relevance of alerts and messages, while upholding strict standards around customer privacy and data protection. Yet, Airwallex knew they had to streamline ETL and pipeline development to accommodate increasing volume and variety across financial, product and unstructured datasets. Timothy Wong, Global Vice President of Data and AI at Airwallex, explained, “We had more teams relying on data for dashboards, customer experiences and automation, but our architecture wasn’t built to scale that demand efficiently. In fintech, the bar is high, so we had to figure out a solution fast.”
When BigQuery and OSS complexity became a bottleneck
Airwallex’s architecture was primarily centered around a structured data warehouse used for BI dashboards and analysis. While this approach worked well early on, limitations became clear as the business scaled, and the demand for faster, more flexible data products increased. Silos made it harder to scale data access and delivery consistently, as different groups built and maintained their own workflows and assets, slowing collaboration and complicating data serving. Not to mention, new initiatives, particularly those tied to customer-facing analytics, automation and AI, required a next-generation architecture that could support broader data types and higher velocity development.
As they introduced additional OSS tooling, costs increased as usage grew, and the overall environment became more complex to maintain. Operating OSS components, such as Airflow and Spark, required significant time from data engineering teams, pulling focus away from higher-impact work like enabling business stakeholders, shipping new data applications and expanding machine learning and AI capabilities within their proprietary platform. To meet fintech requirements for security, compliance and always-on reliability, Airwallex needed a more unified foundation that could democratize access to data without sacrificing control, auditability or operational performance — one they built with Databricks on Google Cloud.
Unifying data serving and governance on Google Cloud
To scale governed analytics and AI, Airwallex chose Databricks to unify data serving and accelerate ML while selecting Google Cloud for their security posture, global scale and Gemini integration. First, Airwallex established a unified data foundation by standardizing on Delta Lake in Databricks as their core storage layer and single source of truth, with Google Cloud Storage (GCS) providing scalable, secure object storage underneath. Using a medallion architecture that classifies data into Bronze, Silver and Gold layers, the team managed and curated a growing mix of financial, product and unstructured data in a consistent way that supported both analytics and AI applications. This approach also helped reduce dependency on a traditional warehouse model and relieve cost pressure from heavy BigQuery usage, as data volumes and business demands continued to increase.
Next, as data use cases expanded across the business, Airwallex needed the next generation of governance in AI that could scale just as quickly. With Databricks Unity Catalog, the team centralized role-based access controls and auditing to ensure the right users could work with the right data at the right time. Combined with Google Cloud’s security-first foundation, this much-needed change supported fintech-grade requirements for security, compliance and reliability while enabling faster innovation. As a result, Airwallex now supports both internal teams and customer-facing analytics with consistent access policies while staying audit-ready as development accelerates. “Databricks and Google Cloud lets us focus our engineering effort on enabling the business instead of tuning infrastructure. With robust governance and strict access controls built in, we can scale new analytics and AI use cases confidently, even as they require access to different data types, without risking over-exposure,” added Timothy.
Speaking of customer-facing analytics, Databricks helped Airwallex serve analytics directly to customers from the lakehouse, while Google Cloud supported secure external delivery requirements. By centralizing data serving across previously siloed departments, Airwallex improved consistency and trust in customer-facing reporting and proprietary data products. These external improvements paved the way for Airwallex to support internal BI and self-service analytics at scale. As the team shifted internal dashboards to the Databricks Platform, they expanded self-service experiences within Databricks. This included using Genie and AI assistants within Databricks to enable non-technical users - including the marketing team - to explore performance data and self-serve metrics such as campaign ROI without needing SQL. For more technical users, Databricks notebooks provided greater flexibility for advanced analysis and development, while integrations with Google Workspace tools such as Google Sheets supported lightweight reporting across stakeholders.
Because of these changes, Airwallex accelerated AI development by giving teams a governed environment to experiment in and ship data products faster. Using Databricks Assistant, in particular, developers moved quickly through tasks like data exploration and coding while keeping access controlled and auditable. Running on Google Cloud also positioned Airwallex to integrate Gemini as part of their AI strategy and maintain flexibility through additional integrations with Anthropic and OpenAI. All in all, the partnership between Databricks and Google Cloud has given Airwallex the ability to expand AI initiatives without adding operational overhead or cost.
Accelerating decision making to improve platform reliability
With Databricks on Google Cloud, Airwallex accelerated how quickly teams can turn data into action while strengthening governance and data accessibility across the organization. “Through our partnership with Databricks and Google Cloud, we achieved 3–4x faster time to insight, which helped us make more consistent decisions as we scaled and delivered new capabilities to our customers,” concluded Timothy.
On the engineering side, Airwallex delivered major efficiency gains by reducing source-to-gold latency from one hour to 15 minutes, supporting their goal of faster, more reliable ETL. They also simplified operations by consolidating 1,600 batch pipelines into 60 declarative pipelines (a 96% reduction) using Lakeflow Spark Declarative Pipelines (SDP). This move dramatically reduced pipeline management overhead and freed engineers to focus on higher-impact work. While Airwallex hasn’t yet tied these improvements directly to revenue yet, they are already seeing clear productivity gains through Databricks Notebooks, Databricks Assistant and Databricks SQL, enabling teams across the business to access and work with data more efficiently.
As they look towards the future, Airwallex plans to deepen their insights-driven approach by scaling Genie across the organization, with plans to build AI agents on Databricks and develop chatbots that synthesize unstructured customer feedback for product managers to release features customers actually want. The team is also expanding machine learning workloads, including growth ML models, and partnering with Databricks to explore new AI use cases, such as observability with logs and evaluating performance across different models. Ultimately, this roadmap supports Airwallex’s mission to remove friction, cost and complexity from global banking.