Talk catalogue
A starting set of talks — each one can be tailored to the audience (research vs. industry, practitioner vs. leadership).
Risk-Based Testing That Actually Ships
From a defect-prediction model to a production microservice that gates pull requests — the design decisions, the calibration trade-offs, and the threats to validity nobody talks about.
The file_age_days Bug
A worked example of disclosure-first empirical software engineering: how a sign bug contaminated 40% of rows, how it was caught, and what changed in the final model.
LLMs + Rule Verifiers for Test Generation
Why deterministic rule verification is not optional, and how symbolic mutation indicators give domain-agnostic adequacy without re-implementing mutation testing.
Cross-Repository Transfer Learning for Defect Prediction
Hands-on with leave-one-repository-out methodology; the AUC–F1 asymmetry; calibration as a deployment criterion.
Self-Healing Locators: Hype vs. Engineering
What works, what doesn't, and why confidence-gated heal-vs-flag matters more than raw heal-rate numbers.
AI Governance for Testing Teams
Prompt versioning, payload controls, author privacy, and audit trails — built into the architecture, not bolted on.