Ali-Reza Adl-Tabatabai describes code review and continuous integration as a centralized quality gate that becomes slower and more expensive as software teams grow. AI increases both the number and size of pull requests, leaving teams to choose between careful but slow review and risky rubber-stamping.
Ali-Reza Adl-Tabatabai outlines an agentic validation workflow that reviews changes, reports important issues, analyzes CI failures and can fix defects until the checks pass. Teams progressively grant stronger permissions, from advisory review to blocking unresolved findings and eventually approving or merging changes under explicit rules.
The architecture separates workflow orchestration, a specialized agent harness and model routing. Ali-Reza Adl-Tabatabai discusses using PR and CI categorization to find recurring operational problems, and combining agent reasoning with taint, control-flow, data-flow and software-composition analysis. Product performance and productivity benefits are presented as the speaker's claims, not independently measured outcomes.
Watch on YouTube
