Shray Sharma: Leveraging AI for Self-Healing Tests: Current Capabilities and Future Prospects
summary
Built a PoC that pulls plain-English Jira test cases, feeds the page DOM into an LLM pipeline (llama and mistro), generates code, then runs it in a Playwright test runner as a verification loop with up to three retries; failing fixes become manual tickets. The system converts accepted output into Playwright scripts via a second model (Nemo), supports Cypress/Selenium style locators, uses context chunking for large DOMs, and reports a 90% auto-fix rate in the demo.
AI-written summary · about
chapters (9)
- 0:00AI in testing motivation
- 3:37Problem statement and basics
- 8:16Outdated info and live data
- 11:11Hallucination and response time
- 14:02Context in AI models
- 18:16Cost reduction and hardware
- 20:26Demo: AI test automation
- 26:23Auto-fixing loop in practice
- 29:32Q&A
related talks
15mLLMs and Test Automation: A New Era of Self Healing
39mHow to Use AI with Playwright for UI & API Testing
48mSupercharged Testing: AI-Powered Workflows with Playwright + MCP
49mReal World Lessons from Scaling Test Automation
43mExploratory Testing with AI
32mSelf-healing system for UI tests using ML
29mPlaying your tests wright
35mState of end-to-end testing for modern web applications