TiDB HTAP for Financial Intelligent Risk Control: Architecture, Challenges, and Solutions

TiDB HTAP can meet the real‑time processing and analysis needs of intelligent risk control scenarios in finance, but it also faces several challenges; this talk summarizes TiDB HTAP's practice in the financial intelligent risk control field.

Financial intelligent risk‑control system overall architecture

Challenges of intelligent risk control in digital transformation

TiDB HTAP intelligent risk‑control solution

TiDB practice in financial risk control

Future outlook and challenges

1. Evolution of the financial risk‑control system – As digital transformation progresses, risk‑control must handle diverse channels (payment platforms, apps, APIs, etc.) and various risk types (pre‑loan credit, mid‑loan delinquency, post‑loan management, payment fraud). Federated learning is emerging to improve data security and model accuracy.

2. Intelligent risk‑control system components – Includes a risk‑algorithm platform for model training, a strategy engine for orchestration and execution, an quota system for unified credit limits, a data‑service engine for real‑time variable computation, and credit/collection systems requiring human interaction.

3. Intelligent risk‑control architecture – Business flows into an entry system, persisting data and providing management. A variable center unifies all risk variables, while data services ingest offline and real‑time sources. The rule engine executes policies based on variables, and operational support (credit review, analytics dashboards) monitors key metrics.

4. Challenges in digital transformation – Real‑time requirements demand second‑level processing; OLTP + OLAP workloads require both ACID consistency and complex analytical capabilities; high‑frequency multi‑point writes cause data fragmentation and increase costs.

5. Industry directions – Build a unified risk data marketplace, simplify complex database stacks, and provide one‑stop real‑time analytics for massive data volumes.

TiDB HTAP solution highlights

Flexible real‑time ingestion via TiDB DM, TiCDC, Flink + Kafka, and connectors for MySQL, Oracle, etc.

HTAP architecture with row‑store TiKV and column‑store TiFlash, supporting MPP queries for massive multi‑table joins in seconds.

Integration with big‑data ecosystems (Spark, Flink, TiSpark) for seamless data processing.

Real‑time data processing engine that chooses optimal storage (row, column, or MPP) based on cost models.

Advantages in intelligent risk‑control metric calculation, enabling second‑level query responses for complex scoring models.

Practical cases demonstrate how TiDB unifies heterogeneous data sources, reduces code complexity, and supports both internet banking and securities firm risk‑control scenarios with real‑time monitoring and alerting.

Future work focuses on building a unified risk monitoring platform, creating a risk data marketplace, and accelerating the financial risk‑control value chain by abstracting technical complexity and monetizing data assets.

Thank you for listening.