A Multi-Source Financial Data Lake, Built at Zero Net Cost

Building financial connectivity infrastructure for businesses across Latin America and beyond, this Miami-based fintech reached a point where fragmented data spread across MongoDB and QuickBooks was limiting the speed and quality of executive decision-making. C-suite leaders shared a clear ambition: consolidate operational data into a governed, queryable analytics layer capable of growing with the business. Rather than waiting for a full-scale data warehouse program, the team chose to move quickly through a phased proof-of-concept approach — validating architecture and business value in weeks, not quarters. Protagona was brought in to design, build, and manage that program end to end.

Across two sequential POCs, the objective was to stand up a governed AWS data lake, integrate MongoDB and QuickBooks as the first two business-critical sources, and deliver a self-service analytics layer through Amazon QuickSight — complete with interactive dashboards and natural-language querying. Each phase was structured to be fully offset by AWS POC funding, making the program cost-neutral while producing production-relevant, reusable architecture the organization could extend independently.

The core technical challenge was not simply connecting two data sources to a lake — it was designing an architecture flexible enough to absorb new sources without rebuilding foundational infrastructure each time. MongoDB and QuickBooks represent fundamentally different data models: one a document store with semi-structured JSON, the other a transactional accounting system with its own API and connector ecosystem. Structuring both into a coherent, SQL-queryable layer that Amazon QuickSight could reliably consume required careful ETL design and schema governance from the outset — decisions made in the first POC had to hold up as the second source was layered in.

‍

On the program side, navigating AWS POC funding mechanics added real complexity. Funding requests, opportunity-stage corrections, and funding category alignment all required active management to ensure approvals landed before project timelines began. Executing two back-to-back funded engagements with minimal gap between them meant managing delivery and funding workflows in parallel, with no margin for administrative delays to derail a three-week build cycle.

Protagona designed and delivered both POCs on a shared architectural foundation. For the first engagement, the team established a governed data lake using AWS Glue Data Catalog and AWS Lake Formation — defining the permission model, storage structure in Amazon S3, and Glue crawler configuration needed to support ongoing source expansion. MongoDB was connected as the initial data source, with AWS Glue handling extraction, transformation, and cataloging of semi-structured document data into an SQL-accessible format. Amazon QuickSight was layered on top, delivering an interactive dashboard and a natural-language query topic through QuickSight Q that allowed business users to ask questions directly against the lake without writing queries.

‍

The second POC extended this foundation rather than replacing it. QuickBooks was integrated using Amazon AppFlow's native connector, with Glue crawlers automatically re-indexing the lake as new financial data arrived. Financial transactions from QuickBooks were tagged to connect to MongoDB records, unifying analytics across both sources. The ETL pipeline was designed for recurring ingestion, with QuickSight's SPICE engine refreshing on a scheduled cadence to keep dashboards current. The architecture was also structured to support row-level security and multi-tenancy in future phases, preserving the ability to scale toward per-client data isolation when the business requires it.

‍

Beyond delivery, Protagona managed the full AWS funding lifecycle for both engagements — including POC funding requests and opportunity-stage alignment — ensuring both projects were approved and executed at zero net cost to the client.