Research

From signals to intelligence.

We develop machine intelligence that understands real systems—their dynamics, uncertainty, anomalies, and human context.

Robotics research
Our vision

One connected intelligence layer for observing, understanding, deciding, and acting in real systems.

ObserveUnderstandDecideAct
01
Research direction

Forecasting, Control & Optimization

Learning how complex systems evolve—and turning that knowledge into better decisions.

We develop probabilistic and deep time-series models that capture uncertainty, missing information, and changing operating conditions. Forecasts are then connected to optimization and learning-based control for energy and industrial applications.

Methods

Probabilistic forecasting

Deep sequence models

Reinforcement learning

Decision-making under uncertainty

Applications

Energy demand

Thermal-hydraulic systems

EV charging

Industrial operations

02
Research direction

Anomaly Detection

Finding rare and consequential signals with less dependence on labeled failures.

We study anomaly detection for tabular, image, vibration, and multivariate sensor data. Our recent work explores the intrinsic representations of untrained neural networks, robust distance-based scoring, and explanations that help people understand why a sample is considered abnormal.

Methods

Untrained neural networks

Representation learning

Explainable AI

Uncertainty & calibration

Applications

Nuclear safety

Predictive maintenance

Machine signals

Visual inspection

03
Research direction

Vision–Language Action

Connecting perception and language understanding to purposeful action in the physical world.

We study multimodal models that ground visual observations and natural-language instructions in robotic actions. Our goal is adaptive embodied intelligence that can understand context, generalize across tasks, and interact safely with real environments.

Methods

Vision–language models

Robot learning

Multimodal representation

Embodied AI

Applications

Adaptive manipulation

Human–robot interaction

Industrial robotics

Autonomous systems

04
Research direction

Augmented Intelligence

Designing AI that extends human awareness instead of hiding the system behind automation.

We combine foundation models, digital twins, robotics, and immersive interfaces to make complex industrial information easier to access and act on. Our focus is the practical integration of machine intelligence with domain expertise and human judgment.

Methods

Machine-signal foundation models

Large language models

Digital twins

Human–AI interaction

Applications

Industrial copilots

AR/VR interfaces

Robot intelligence

Operator support

NextExplore funded projects