Session

Beyond Model Accuracy: Evidence-Gated AI for Reliable Industrial Material Inspection

In industrial material inspection, high model accuracy does not necessarily guarantee reliable decisions. What happens when an AI system produces a confident prediction but lacks sufficient visual evidence to support it?

This lightning talk introduces SecureCoating-Vision, a research prototype exploring evidence-gated AI for electrode coating inspection. By combining visual detection, segmentation, and evidence-based decision logic, the system distinguishes between inspection results that have sufficient supporting evidence and those that should be held for further review.

Through a practical inspection example, the session explores uncertainty handling, conservative decision-making, and lessons learned from developing AI-assisted inspection workflows.

The central message: reliable industrial AI is not just about predicting accurately, but also about knowing when not to make an automated decision.


Selected session: APAC AI for Science DevFest 2026, organized by Google Researcher Groups.

Format: 10-minute Lightning Talk / Demo.
Event dates: November 27–28, 2026.
Target audience: AI/ML engineers, applied AI researchers, computer vision practitioners, and industrial AI enthusiasts.
Technical level: Intermediate. No special software or equipment required.
The presentation covers an experimental research prototype, not a production deployment.

Hoang Tu Trinh

Software Engineer (Backend, in Training) at Grab | Hackathon Hunter & AI Researcher

Ho Chi Minh City, Vietnam

Actions

Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.

Jump to top