Big Data Implementation Practices and Architecture in a Foreign Bank

The presentation outlines a foreign bank's experience building a big data platform to support its expanding retail and cross‑border services. It begins with the background and objectives, noting a strategic shift in 2017‑2018 that caused a surge in data volume and variety, prompting a need for stronger data quality control, open data capabilities, and robust data development management.

The overall plan consists of five major goals: establishing a bank‑wide unified data architecture, enhancing data openness for business users, adopting storage‑compute separation, implementing practical data governance, and strengthening data security controls. These goals are realized through a layered data architecture (source, ingestion, storage/processing, service bus, and application layers) and a complementary technical stack that integrates Greenplum, Hadoop, Kafka, and other tools.

Implementation is divided into three phases: (1) 2018‑2019 – foundation with an MPP + Hadoop hybrid architecture and Greenplum for regulatory reporting; (2) 2023 – Hadoop platform expansion after a new core system launch, focusing on retail risk management and marketing; (3) 2024 – real‑time applications such as payment monitoring, anti‑fraud, and a management dashboard.

Key practical insights include: sharing a common technical foundation with logical layering, conducting data lineage analysis and impact assessment, defining comprehensive data quality rules (technical legality and business reasonableness), and establishing a full‑lifecycle security and audit framework. The bank also explores AI‑enabled capabilities like NL2SQL‑based BI reporting, LLM‑driven customer‑service knowledge bases, and tag‑based marketing automation.

The final section reviews the outcomes: a solid data foundation that improves core capabilities, supports AI‑driven analytics, and enables secure, governed data usage. A Q&A session addresses scaling considerations, digital transformation challenges for small‑to‑mid‑size banks, and measures to ensure data availability.