1K+ tables. 12 databases. One governed lakehouse. Delivered in 7 months.

How Cuedo built the enterprise data platform a leading Indian NBFC now runs its entire business on.

📁
95%
Work
automation
☑️
12
Source
databases unified
1K
Tables
consolidated
📊
25
ML
models in production

A leading Indian NBFC, operating across vehicle finance, home loans, business loans, and insurance, was running its analytics estate across Apex, Power BI, and Excel based MIS, with no centralised integration layer. Every new dashboard or model meant analysts rebuilding ETL and data marts from scratch. The same metric had different definitions depending on which tool produced it. Decisions were slowed, and trust in the numbers was eroding.

12
Source databases
no integration layer
1,000
Tables
in silos
4
Disconnected systems
(Oracle, MySQL, MS SQL, PostgreSQL)
0
Centralised governance
or single source of truth

What Cuedo Built

An end to end enterprise data platform on Google Cloud, anchored on BigQuery with a Medallion (Bronze, Silver, Gold) architecture, delivered through five configurable frameworks, each independently deployable via its own CI/CD pipeline.

01 On-Prem Sources (Oracle, MySQL, MS SQL, PostgreSQL)
02 Dataflow / Datastream / Dataproc
03 Bronze Layer
04 Silver (Cloud Composer, Type 2 SCD)
05 Gold (Shared KPIs & Datamarts)
06 Power BI / Vertex AI

Key engineering decisions:

  • Medallion architecture in BigQuery, with Bronze for raw replication, Silver for cleansed entity models, and Gold for governed, business ready datasets
  • Five configurable frameworks, Data Ingestion, Ingestion Orchestration, Data Model, Data Transformation, and Data Orchestration, each with independent CI/CD and rollback
  • Dataflow and Datastream for batch, incremental, and real time CDC ingestion; Dataproc for complex unstructured and UDT heavy sources
  • Full Infrastructure as Code via Terraform, inside a hardened GCP Landing Zone with VPC, IAM, CMEK encryption, and SIEM integration
  • Vertex AI powering MLOps for 25 production models, including automated training, deployment, and drift monitoring
  • Dataplex for catalog, lineage, classification, and policy enforcement across the entire estate
  • Config driven ingestion templates, meaning new tables onboard by editing configuration, not writing new pipelines
Before
Inconsistent KPIs across reports
No centralised data governance
Fragile legacy pipelines
After
700+ governed metrics, one definition each
Full lineage, RBAC, and masking via Dataplex
DR aware, zero manual edit failover

1,000+ tables. 12 databases. Delivered in 7 months. A platform giving the business one governed answer to every KPI question, powering 25 ML models in production, with the client's own team trained to run and extend it after go-live.