Research

Explorations across adaptive computation, AI systems, sensing, and health-tech.

Liquid Computation

Adaptive Dynamics for Real-World Systems

We investigate liquid-style computation and dynamical models for systems that must adapt continuously to changing inputs and environments.

Continuous-time / stateful models for streaming signals
Robustness under distribution shifts
Energy-efficient inference and deployment constraints

Applications of AI Models

From Prototypes to Real Deployments

We explore practical workflows for applying modern models to real operational problems, with a focus on reliability and evaluation.

Model selection and benchmarking
Task-specific prompts, agents, and tool-use
Safety, failure modes, and monitoring

Data Collection, Training, and Testing

End-to-End ML Lifecycle

We develop repeatable pipelines for dataset creation, labeling, training, and evaluation, aimed at minimizing drift and maximizing reproducibility.

Data acquisition and quality controls
Training workflows, ablations, and experiment tracking
Test sets, stress tests, and deployment checks

Object Tracking

Tracking in 2D/3D and Mixed Reality

We study robust tracking systems for real-time applications, including sensor fusion and XR experiences.

Detection-to-tracking pipelines
Latency, stability, and occlusion handling
Tracking evaluation and failure diagnostics

Health-Tech Devices

Sensing, Signals, and Practical UX

We explore device concepts that combine sensing hardware with software intelligence, focusing on usability, privacy, and practical deployment.

Sensor data processing and on-device inference
Reliability, calibration, and user feedback loops
Privacy-first data handling considerations