Ensuring Observability and Explainability in Multimodal Data Pipelines for AI Development

Presented at OXAI’24, part of CASCON 2024.

In this talk, I explored practical strategies for embedding observability in large-scale AI data transformation pipelines. Using a case study focused on multimodal data sources like social media, I discussed key methods for ensuring transparency and accountability in data handling. Highlights included:

• Techniques for monitoring data usage and maintaining data transparency
• Methods to ensure semantic coherence across multimodal datasets
• Insights on integrating human oversight to improve precision and regulatory compliance

OXAI’24 brought together experts and practitioners in the field, fostering an exchange of ideas around creating accountable, explainable AI systems. This event underscored the importance of observability and explainability in enhancing trust and efficiency in AI deployments.