RESEARCH AREAS

I work on Efficient and Explainable artificial intelligence (XAI), World Models (WM) for electromagnetics/wireless, and Learning-based Optimization (L2O) for real-world communication and network decisions. Our goal is interpretable, editable, and deployable methods that connect physics, data, and decision making.

Theoretical Part

Sparse and Efficient AI

Structured sparsity, low-rank, and tensorization across data, features, models, and gradients, with unified compression and distillation for edge–cloud collaboration and federated training. HW/SW co-acceleration (CPU, GPU, FPGA) enables reliable on-device deployment in wireless applications.

Structured sparsity and low-rank modeling

- Compression and distillation for models and gradients Edge/federated learning
- Communication-efficient training and adaptation System efficiency
- Algorithm–system co-design for throughput and energy gains

Explainable AI (XAI)

Characterize generalization bounds and convergence properties of AI methods using high-dimensional statistics and nonconvex optimization, with an emphasis on interpretability and reliability for wireless applications.

Application Part

World Models (WM) for Radio Environment

Physically grounded, data-driven models of radioEM environments that fuse geometry and materials (e.g., BIM, point clouds, maps), contextual signals (vision, language, IMU, GNSS), and RF measurements (sweeps arrays). We combine neural representations (3DGS and NeRF variants, diffusion and energy models, GNNs, neural operators such as FNO, DeepONet, and GNO) with PDE and boundary consistency, constitutive relations, and structured sparsity or low-rank priors. The models support editable inversion, uncertainty quantification, and active re-measurement for localized statistical channel modeling and environment-aware communication.

Localized statistical channel modeling

Neural radio radiance fields; structured Bayesian inference on graphs Physics-informed consistency

PDE and boundary consistency; material priors; calibrated uncertainty Multimodal fusion

Geometry/materials plus RF arrays plus contextual signals

Learning-based Optimization (L2O)

Differentiable, learning-augmented solvers (RL, GNN, and neural combinatorial optimization) operating on WM for constrained decisions, with feasibility guarantees, learned warm-starts, and fast cross-scenario adaptation.

Communication and RAN optimization

Site placement; beam, power, and spectrum allocation; routing and formation planning Learned solvers

Relax–optimize–and–sample; neural projections; constraint satisfaction Deployment

Edge and federated execution for real-time, certifiable decisions

FieldMind: From Representation to Decisions

FieldMind is our instantiation of WM plus sparse and efficient AI plus L2O, coupled with algorithm–hardware co-design (C, CUDA, FPGA; edge and federated) for online decision making. It maps structure across data, algorithms, and architectures into solvers and hardware for