Research

Applied optimization

Energy Systems

Energy trading and markets, power distribution and management, resource allocation, and distributed optimization in microgrids.

We develop optimization frameworks for modern energy infrastructures — from market clearing and dispatch to resilient microgrid coordination — combining distributed algorithms with realistic operational constraints.

Models of vision

Computational Neuroscience

Optimal models of the human visual pathway to better understand the brain and visual cortex.

Our neuroscience track builds computationally grounded models of visual processing, linking optimization theory with empirical structure of sensory pathways.

Digital Twin

Digitalization of Industry

Digital twins optimize industrial processes by creating a continuously updated virtual replica of physical operations, enabling manufacturers to simulate, predict, and refine workflows without disrupting actual production.

In high-mix/low-volume manufacturing, integrating digital twins with value stream mapping and simulation allows companies to identify bottlenecks, balance workloads, and streamline material flow, resulting in faster throughput and shorter cycle times. In lighthouse factories, end-to-end digital twins combined with artificial intelligence applications compress lead times, accelerate time-to-market, and boost overall productivity by enabling real-time visibility and rapid adjustment of production parameters. Beyond discrete manufacturing, digital twins optimize energy consumption in smart factories by using real-time sensor data to adaptively control systems, reducing waste and improving data integrity across the production network. In complex industries such as shipbuilding, AI-driven digital twin models for nesting and segmentation optimization raise material utilization and shorten manufacturing cycles by testing layout and scheduling alternatives virtually before committing physical resources. Fault detection is another critical capability, as digital twins continuously compare real-time operational data against expected behavior to identify anomalies, deviations, and early signs of equipment degradation, allowing maintenance teams to intervene before minor issues escalate into unplanned downtime. The core mechanism is simulate first, then act—testing new settings, layouts, and schedules in the virtual environment before implementing them on the factory floor, thereby avoiding downtime, material waste, and costly late-stage rework while continuously improving process efficiency over time.