Research Areas

Overview

The Supply Chain Data Science and AI (SCDS) Lab works on applying machine learning, data science, and optimization to problems in manufacturing, logistics, sustainability, and autonomous systems. Our research is motivated by practical challenges in real-world systems, where data is messy, environments change over time, and decisions often have to be made under uncertainty.

We are especially interested in how modern AI systems behave when conditions are imperfect — for example, when sensors degrade, data distributions shift, or the system is operating outside the situations it was trained for. Our goal is to develop AI methods that are not only accurate, but also reliable, interpretable, and safe to use in real applications.

Research Themes

Reliability and Uncertainty in Machine Learning: Many machine learning models perform well on benchmarks but become unreliable when deployed in the field. A core focus of our lab is understanding and managing this gap.

We study:

• How to estimate confidence and uncertainty in deep learning models
• How to detect when a model is likely to fail or become unreliable
• How to design simple rules that prevent high-risk actions when confidence is low

This work is especially important for safety-critical applications such as autonomous monitoring, surveillance, and early warning systems.

Learning Structure and Patterns from Data: We use both supervised and unsupervised learning methods to discover patterns in complex data. This includes work on:

• Fabric structure and defect classification
• Pattern discovery in manufacturing and inspection data
• Anomaly detection in sensor and supply chain data

Our aim here is not only prediction, but also understanding — using data to reveal structure and relationships that are difficult to capture with manual rules.

Data-Driven Experimentation and Evaluation: We are interested in how machine learning can guide data collection and testing, not just analysis. This includes::

• Active learning and adaptive data collection
• Simulation and stress-testing of AI systems
• Systematic evaluation of performance under changing or degraded conditions

This allows us to better understand where systems fail and how to improve them.

Application Areas

Manufacturing and Materials: We work on computer vision and data-driven tools for:

• Textile defect detection and quality control
• Fabric and material structure recognition
• Supporting circular manufacturing and sustainable materials

This work supports more efficient and sustainable production processes.

Autonomous and Embedded Systems: We develop and deploy machine learning models on embedded and edge platforms for tasks such as:

• Fire detection and early warning
• Autonomous monitoring and surveillance
• Operation under limited compute and challenging environmental conditions

This includes model optimization for small devices and robustness to noise, weather, and sensor limitations.

Geospatial and Climate-Related Systems: We combine geospatial data, machine learning, and optimization to study:

• Supply chain resilience under climate and geopolitical disruption
• Risk-aware planning and resource allocation
• Environmental and sustainability impacts

Overall Vision

The long-term goal of the lab is to build AI systems that can be trusted in practice — systems that understand when they are uncertain, behave conservatively when risk is high, and support better human and organizational decision-making.