projects
A growing collection of your cool projects.
work
5G Testbed and Use-case Deployment
Containerized and automated the deployment of a fully functional 5G testbed integrating srsRAN and Open5GS, enabling experimentation and validation of intelligent algorithms. Also deployed cloud-gaming and VR use-cases over the 5G testbed.
Built AI-based digital twins of the 5G network for accurate, scalable and generalizable network modeling. Designed composable and differentiable per-VNF models to support gradient-based optimization workflows.
- vChainNet: Accurate and Scalable End-to-End Slice Modeling for 5G and Beyond Networks — WINCOM 2026
- vNetRunner: Per-VNF Slice Modeling for 5G and Beyond Networks — IEEE/IFIP NOMS 2025
- MicroOpt: Model-driven Slice Resource Optimization in 5G and Beyond Networks — IEEE TNSM 2025
Dynamic Resource Management
Reinforcement learning and primal–dual optimization for end-to-end dynamic resource allocation in 5G networks, ensuring efficient resource utilization under QoS/SLA constraints.
- Generalizable Resource Scaling of 5G Slices using Constrained Reinforcement Learning — IEEE/IFIP NOMS 2023
- MicroOpt: Model-driven Slice Resource Optimization in 5G and Beyond Networks — IEEE TNSM 2025
AI-Native Link Adaptation
Designed and validated several AI-algorithms for dynamic link adaptation in 5G networks, enabling real-time modulation and coding scheme (MCS) selection under dynamic wireless conditions. Leveraged advanced deep learning algorithms to outperform current approachs for enhanced spectral efficiency.
Adaptive Network Monitoring
Deployed a big-data processing pipeline and Engineered an adaptive cloud-native monitoring framework for slice-level KPI monitoring that dynamically adjusts data collection granularity to reduce monitoring overhead
- MonArch: Monitoring Architecture for 5G and Beyond Network Slices — IEEE TNSM 2024
5G Service Orchestration
Developed intelligent orchestration mechanisms for automating slice admission control and VNF embedding in 5G networks. Leveraged multi-agent deep reinforcement learning (MADRL), graph neural networks, and constrained optimization to improve scalability, fairness, and QoS guarantees.
- Data-driven Online Slice Admission Control and Resource Allocation for 5G and Beyond Networks — arXiv 2025 (under review at INFOCOM ’26)
- Multi-Agent Deep Reinforcement Learning for Slicing and Admission Control in 5G C-RAN — IEEE/IFIP NOMS 2022
- Coordinated Slicing and Admission Control Using Multi-Agent Deep Reinforcement Learning — IEEE TNSM 2022
- Generalizable 5G RAN/MEC Slicing and Admission Control for Reliable Network Operation — IEEE TNSM 2024
CSI-Based Activity Recognition
Device-free human activity recognition with commodity Wi-Fi/USRP CSI and deep learning; applications in smart environments, healthcare, and security.
- True Detect: Deep Learning-based Device-Free Activity Recognition using WiFi — IEEE WCNC Workshops 2020