Research

My research lies at the intersection of networking, learning, and systems for intelligent decision-making. I am especially interested in how constrained devices, edge servers, and cloud resources can be coordinated to deliver practical and reliable AI.

Hierarchical and Cascade AI

Modern AI systems often need to decide whether to answer locally or to offload computation to a stronger edge or cloud model. My work studies the algorithmic foundations of hierarchical inference, including confidence-aware routing, decision modules, queueing effects, and performance tradeoffs across the device-edge-cloud hierarchy.

Inference Offloading and Edge Systems

I study algorithms that determine when to execute models on-device, when to use an edge server, and when to escalate to the cloud. The goal is to jointly optimize latency, communication, energy, accuracy, and monetary cost rather than treating them as separate design dimensions.

Age of Information and Freshness

Another focus of my work is information freshness in distributed sensing and monitoring systems. I am interested in how stale information affects control, estimation, and decision-making in networked environments.

Online Learning and Networked Decision Systems

I also work on online learning, stochastic optimization, queueing, and decision-making problems that arise in adaptive systems. These tools support learning-based routing, hierarchical inference, and broader resource allocation in networks and information systems.

Current Themes

  • Reliable and cost-efficient Edge AI for real deployments.
  • Routing and orchestration across heterogeneous models.
  • Freshness-aware sensing and monitoring at scale.
  • Learning and control under communication and compute constraints.

Past Research

Age of Information Analysis and Optimization

Age of Information (AoI) tracks how old the freshest delivered update is, so it reflects both sampling frequency and communication delay. This work derived exact AoI distributions for basic queueing models and designed sampling and transmission policies that keep information fresh.

  • Statistical Age-of-Information Bounds for Parallel Systems: When Do Independent Channels Make a Difference?, IEEE JSAIT 2023.
  • Statistical Guarantee Optimization for AoI in Single-Hop and Two-Hop FCFS Systems with Periodic Arrivals, IEEE TCOM 2021.
  • Minimum Achievable Peak Age of Information Under Service Preemptions and Request Delay, IEEE JSAC 2021.
  • On the Distribution of AoI for the GI/GI/1/1 and GI/GI/1/2* Systems: Exact Expressions and Bounds, IEEE INFOCOM 2019.

Transient Delay Analysis and Optimization

For time-critical industrial traffic that arrives in bursts, steady-state results can be misleading. Using stochastic network calculus, this work bounded end-to-end delay violation probabilities given the transient queue state and used them for resource allocation in wireless protocols such as WirelessHART.

  • Scheduling of Wireless Edge Networks for Feedback-Based Interactive Applications, IEEE TCOM 2022.
  • Transient Analysis for Multi-hop Wireless Networks Under Static Routing, IEEE/ACM Transactions on Networking 2020.

Computation Offloading Algorithms (PhD Work)

Edge and fog computing place compute one hop from the user. During my PhD, I designed semi-online and restart-based algorithms with provable competitive ratios for offloading tasks from mobile devices to edge servers.

  • Single Restart with Time Stamps for Parallel Task Processing with Known and Unknown Processors, IEEE TPDS 2019.
  • Efficient Minimization of Sum and Differential Costs on Machines with Job Placement Constraints, IEEE INFOCOM 2017.
  • Semi-Online Algorithms for Computational Task Offloading with Communication Delay, IEEE TPDS 2016.