Mr. Karan Alang

AI That Never Sleeps: Building Adaptive Data Pipelines for an Evolving World Meta-Learning Approaches for Continual Adaptation in Non-Stationary Streams

Abstract:

As artificial intelligence becomes embedded in critical domains such as healthcare, finance, and autonomous systems, one challenge constantly undermines long-term reliability: data drift. Real-world data streams are dynamic, noisy, and unpredictable—causing traditional machine learning pipelines to degrade over time.

This keynote introduces Meta-Learned Data Pipeline Adaptation (ML-DPA), a novel framework that goes beyond model retraining to dynamically reconfigure the entire data pipeline. By applying meta-learning, ML-DPA enables preprocessing, feature engineering, and normalization strategies to adapt in real-time to evolving data distributions.

Through case studies in autonomous driving, fraud detection, and real-time health monitoring, we will explore how adaptive pipelines mitigate concept drift, reduce catastrophic forgetting, and maintain robust performance in production environments. The talk will also highlight the future of self-adaptive AI infrastructure, where models and pipelines behave more like living systems—constantly learning, adjusting, and improving.

Profile:

Karan Alang is a Principal Software Engineer, Tech Lead, and Architect with 25 years of experience designing and scaling distributed systems, cloud-native platforms, and AI/ML solutions. Throughout his career, he has led innovation and engineering excellence at world-class organizations such as Apple, Versa Networks, 3M, Lenovo, and US Steel — driving impactful work in big data, real-time analytics, and intelligent automation.

He leads the Data Engineering team at Versa Networks Inc., where he architects scalable data pipelines and platforms that support advanced machine learning and security intelligence.

Beyond his technical contributions, he is deeply engaged with the global tech community. He is an IEEE Senior Member, ACM member, SCRS Fellow, a Member of the Forbes Technology Council, staying closely connected with emerging research and thought leadership in the field.

He is also an open-source contributor to the Apache Airflow community and actively supports innovation in data orchestration. In addition, he has served as a hackathon judge and technical speaker at multiple industry events.