System 05Applied AI Research2019—2023Research

Oil & Gas Leak Computer Vision

A computer-vision pipeline for oil and gas leak detection built during a four-year AI research fellowship.

94%Detection precision
Context

Industrial leak detection required an applied imaging pipeline capable of extracting a reliable signal from a large visual dataset.

Intervention

Designed and evaluated a computer-vision pipeline across more than 10,000 images.

My role

Lead research fellow, model development, experimentation, evaluation, and publication.

Research scope

The project explored computer imaging for oil and gas leak detection. It combined data preparation, model experimentation, evaluation, and research communication in a multi-year fellowship environment.

Evidence

  • More than 10,000 images processed by the pipeline.
  • 94% reported precision.
  • Peer-reviewed AI and computer-vision research presented at conferences.
  • Work completed as part of a fellowship awarded to a top-5% Informatics student.

Foundation for production work

The research built the evaluation discipline that later informed production ML work: define the metric, understand the failure mode, and treat model output as one part of a larger system.

Next system · 01

AI Gateway