Mr. Krishnam Raju Narsepalle
Data Engineering for AI-Driven Computer Vision: Challenges and Solutions
Abstract:
The integration of artificial intelligence into computer vision applications is redefining how machines interpret and act on visual data. However, scaling these systems requires more than just powerful models—it demands robust, efficient, and adaptable data engineering. This talk explores the critical role of data engineering in enabling real-time and batch processing for AI-driven computer vision, focusing on challenges such as data volume, annotation quality, model drift, and end-to-end pipeline automation. Drawing from enterprise experience across finance, retail, and telecom domains, this session highlights practical solutions using modern cloud-native platforms like Google Cloud Platform (GCP) and AWS. Technologies such as Apache Kafka, MongoDB, Google Dataflow, and BigQuery will be discussed in the context of image data ingestion, feature extraction, and deployment at scale. The session will also address MLOps integration, edge-computing considerations, and how to bridge the gap between raw image streams and intelligent decision-making systems, delivering actionable insights to both practitioners and researchers in computer vision.