What if AI generated your test data?
summary
ING uses an MIT‑origin open‑source engine (now DataSibo) to train on ~10–20k real rows, build a statistical model and sample 10k SEPA payments in less than 1 minute for acceptance testing; models respect column distributions and correlations, allow conditional and out‑of‑range sampling to force rare cases, and generate thousands of request/response pairs for WireMock request/response generation. PII is handled by fakers or learned synthetic formats, models train in minutes, and integration targets include Azure pipelines and sharing synthetic data (not the models) with external parties.
AI-written summary · about
chapters (12)
- 0:00Introduction and setup
- 3:08Project background and motivation
- 8:46Why synthetic test data
- 13:24How synthetic data works
- 20:14Tools and algorithms used
- 27:04First real-world example
- 32:25Handling new systems
- 34:33Privacy and personal data
- 37:24Outliers and edge cases
- 41:18Testing APIs with mocking
- 45:14Open source and resources
- 46:42Q&A session
related talks
38mRedefining Test Data Strategy in the Gen AI Era
55mAI-First Testing
32mQuality engineering in the AI age: From testing AI to leveraging GenAI
43mAI-Generated API End-to-End Tests: A Legacy Rewrite Case Study
01h 00mHow ANZ Leverages GenAI to Transform Testing Activities
01h 04mImpact of AI on #SoftwareTesting: Are Testers Ready? 🔮
44mAI Empowered Software Testing
42mIt's Inevitable, Mr. Anderson: Machines, Humans, and the Future of Testing