Publications

(2026). Should SLMs be in the Driver's Seat? A Study of Local Models as Interpreters in Cloud-Written Agent Programs. The 1st NeurIPS Workshop on SLMs for Agentic Systems (SLM-Agents'26) To appear..

(2026). fRAC: Remote Accelerator Calls on Hostless FPGAs. SIGOPS-ATC'26.

PDF DOI Technical Report Artifact (Available, Functional, Reproduced)

(2026). Exploring Tree-Based Approaches for Collective Communications on Cerebras WSE-3. SC'26.

PDF

(2026). DynSL: Resource-Adaptive Model Training across Heterogeneous Edge Devices. SoCC'26.

PDF DOI

(2026). COMED: The Missing Middle Between Routing and Collaboration in Multi-LLM Inference. AACL-IJCNLP'26.

PDF

(2026). When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs. COLM'26 To appear..

Technical Report

(2026). MCSched: Memory-Controller-Aware Scheduling for Embodied LLM Workloads on NVIDIA Jetson. APSys'26.

PDF

(2026). Congestion-Aware Serving of Agentic LLM Applications. APSys'26.

PDF

(2026). κ-LoRA: Condition Numbers Reveal Which LoRA Matrices Worth Updating. arXiv:2607.22489.

PDF

(2026). Label-free Dataset Profiling for Federated Client Clustering. Data Mining and Knowledge Discovery, 40.

PDF DOI

(2026). FOCUS: DLLMs Know How to Tame Their Compute Bound. ICML'26.

PDF Technical Report

(2026). PICO: Performance Insights for Collective Operations. ISC High Performance'26.

PDF

(2026). Peeking Under the Hood of Multi-Agent Systems. CAIS'26.

PDF

(2026). Opportunistic Telemetry Transport in Hardware-Accelerated Observability Pipelines. NetCompute'26.

PDF

(2026). Observability Is Eating Your Cores: Fine-Grained Analysis of Microservice Metrics with IPU-Hosted Sketches. NSDI'26.

PDF Code Artifact (Available, Functional, Reproduced)

(2026). Cost-effective and Reliable Global Internet Peering with Programmable Switches. NSDI'26.

PDF

(2026). Application-Level Function Pipelines on In-Network FPGA Accelerators. IFIP Networking'26.

PDF

(2026). Why Smaller Is Slower? Dimensional Misalignment in Compressed LLMs. EuroMLSys'26.

PDF

(2026). Reducing the GPU Memory Bottleneck with Lossless Compression for ML. EuroSys'26.

PDF Code Artifact (Available, Functional, Reproduced)

(2026). PETS: Inference-Time Differentially Private Synthetic Time Series Generation. The 1st ICLR Workshop on Time Series in the Age of Large Models (TSALM'26).

PDF

(2026). Nonparametric Teaching of Attention Learners. ICLR'26.

PDF

(2026). RAP: KV-Cache Compression via RoPE-Aligned Pruning. arXiv:2602.02599.

PDF

(2026). Efficient and Adaptable Detection of Malicious LLM Prompts via Bootstrap Aggregation. arXiv:2602.08062.

PDF

(2026). MAESTRO: Multi-Agent Evaluation Suite for Testing, Reliability, and Observability. arXiv:2601.00481.

PDF

(2025). A Network Arena for Benchmarking AI Agents on Network Troubleshooting. arXiv:2512.16381.

PDF

(2025). Toward an AI-Native Internet: Rethinking the Web Architecture for Semantic Retrieval. arXiv:2511.18354.

PDF

(2025). ACING: Actor-Critic for Instruction Learning in Black-Box LLMs. EMNLP'25.

PDF

(2025). Quantize Once, Train Fast: Allreduce-Compatible Compression with Provable Guarantees. ECAI'25.

PDF Technical Report

(2025). FPGAs Are the Hero In-Network Computing Needs. APSys'25.

PDF DOI

(2025). Flashback: Understanding and Mitigating Forgetting in Federated Learning. FLTA'25.

PDF Technical Report

(2025). DMAS-Forge: A Framework for Transparent Deployment of AI Applications as Distributed Systems. The 1st Workshop on Systems for Agentic AI (SAA'25).

PDF

(2025). Cloud abstractions for AI workloads. APSys'25.

PDF DOI

(2025). Between Promise and Pain: The Reality of Automating Failure Analysis in Microservices with LLMs. APSys'25.

PDF DOI

(2025). Scaling SCIERA: A Journey Through the Deployment of a Next-Generation Network. SIGCOMM'25.

PDF

(2025). Latent Inference for Effective Multi-Agent Reinforcement Learning under Partial Observability. The 18th European Workshop on Reinforcement Learning (EWRL'25).

PDF

(2025). LLM Optimization Without Data Sharing: A Split Learning Paradigm. MetaCom'25.

PDF

(2025). Idle No More: Boosting Distributed Pipeline Training via FluidPipe. HotDiML'25.

PDF

(2025). Analysis and Tuning of Knowledge Distillation for Efficient Collaborative Learning. ICDCS'25 (short paper).

PDF Technical Report

(2025). Assassyn: A Unified Abstraction for Architectural Simulation and Implementation. ISCA'25.

PDF

(2025). Decentralized Personalized Federated Learning. AISTATS'25.

PDF

(2025). Query-based Knowledge Transfer for Heterogeneous Learning Environments. ICLR'25.

PDF

(2025). Towards a Unified Framework for Split Learning. EuroMLSys'25.

PDF

(2025). OptimusNIC: Offloading Optimizer State to SmartNICs for Efficient Large-Scale AI Training. EuroMLSys'25.

PDF

(2025). Global-QSGD: Allreduce-Compatible Quantization for Distributed Learning with Theoretical Guarantees. EuroMLSys'25.

PDF

(2024). Where is the Testbed for my Federated Learning Research?. SEC'24.

PDF

(2024). Review and Comparative Evaluation of Resource-Adaptive Collaborative Training for Heterogeneous Edge Devices. ACM Transactions on Modeling and Performance Evaluation of Computing Systems.

PDF DOI

(2024). Protecting Confidentiality, Privacy and Integrity in Collaborative Learning. arXiv:2412.08534.

PDF

(2024). Continuous Exact Explanations of Neural Networks. ICDM'24.

PDF

(2024). Immediate Communication for Distributed AI Tasks. The 2nd Workshop on Hot Topics in System Infrastructure (HotInfra'24).

PDF

(2024). FilFL: Client Filtering for Optimized Client Participation in Federated Learning. ECAI'24.

PDF Technical Report

(2024). Towards a Flexible and High-Fidelity Approach to Distributed DNN Training Emulation. APSys'24.

PDF

(2024). Collaborative Learning Experimentation Testbed. APSys'24.

PDF

(2024). SqueezeNIC: Low-Latency In-NIC Compression for Distributed Deep Learning. NAIC'24.

PDF

(2024). Proof-of-Concept of a Flexible and High-Fidelity Approach to Distributed DNN Training Emulation. NAIC'24.

PDF

(2024). OmNICCL: Zero-cost Sparse AllReduce with Direct Cache Access and SmartNICs. NAIC'24.

PDF

(2024). Train your cake and eat it too! Repurposing collaborative training to tailor LLMs to private data without sharing. The 2nd ICML Workshop on Efficient Systems for Foundation Models (ES-FoMo'24).

PDF

(2024). Towards LLM-Assisted System Testing for Microservices. AI-DCS'24.

PDF

(2023). On Detecting Biased Predictions with Post-hoc Explanation Methods. SAFE'23.

PDF

(2023). MicroView: Cloud-Native Observability with Temporal Precision. CoNEXT Student Workshop'23.

PDF

(2023). Kimad: Adaptive Gradient Compression with Bandwidth Awareness. DistributedML'23.

PDF

(2023). TENSOR: Lightweight BGP Non-Stop Routing. SIGCOMM'23.

PDF

(2023). A Comprehensive Empirical Study of Heterogeneity in Federated Learning. IEEE Internet of Things Journal, 10(16).

PDF DOI

(2023). SAGE: Software-based Attestation for GPU Execution. USENIX-ATC'23.

PDF Code Artifact (Available, Functional)

(2023). With Great Freedom Comes Great Opportunity: Rethinking Resource Allocation for Serverless Functions. EuroSys'23.

PDF Code Artifact (Available, Functional)

(2023). REFL: Resource-Efficient Federated Learning. EuroSys'23.

PDF Code Artifact (Available, Functional)

(2023). A First Look at the Impact of Distillation Hyper-Parameters in Federated Knowledge Distillation. EuroMLSys'23.

PDF

(2023). In-Network Aggregation with Transport Layer Transparency for Distributed Training. ASPLOS'23.

PDF

(2022). Direct Nonlinear Acceleration. EURO Journal on Computational Optimization, 10.

PDF DOI

(2022). Natural Compression for Distributed Deep Learning. MSML'22.

PDF

(2022). Unlocking the Power of Inline Floating-Point Operations on Programmable Switches. NSDI'22.

PDF Project

(2022). RDMA is Turing complete, we just did not know it yet!. NSDI'22.

PDF Code

(2022). Empirical Analysis of Federated Learning in Heterogeneous Environments. EuroMLSys'22.

PDF

(2022). Renaissance: A self-stabilizing distributed SDN control plane using in-band communications. Journal of Computer and System Sciences, 127.

PDF DOI

(2021). Rethinking gradient sparsification as total error minimization. NeurIPS'21.

PDF Code

(2021). Tardis: A Fault-Tolerant Design for Network Control Planes. SOSR'21.

PDF

(2021). LineFS: Efficient SmartNIC Offload of a Distributed File System with Pipeline Parallelism. SOSP'21.

PDF Code Video Artifact (Available, Functional, Reproduced)

(2021). Efficient Sparse Collective Communication and its application to Accelerate Distributed Deep Learning. SIGCOMM'21.

PDF Code Project Technical Report Artifact (Available, Functional, Reproduced)

(2021). GRACE: A Compressed Communication Framework for Distributed Machine Learning. ICDCS'21.

PDF Code Project Technical Report

(2021). DC2: Delay-aware Compression Control for Distributed Machine Learning. INFOCOM'21.

PDF Project

(2021). AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly. ICLR'21.

PDF Code

(2021). Analyzing Learning-Based Networked Systems with Formal Verification. INFOCOM'21.

PDF

(2021). Towards Mitigating Device Heterogeneity in Federated Learning via Adaptive Model Quantization. EuroMLSys'21.

PDF

(2021). Scaling Distributed Machine Learning with In-Network Aggregation. NSDI'21.

PDF Code Project Technical Report

(2021). An Efficient Statistical-based Gradient Compression Technique for Distributed Training Systems. MLSys'21.

PDF Code Project

(2020). Huffman Coding Based Encoding Techniques for Fast Distributed Deep Learning. DistributedML'20.

PDF

(2020). Assise: Performance and Availability via Client-local NVM in a Distributed File System. OSDI'20.

PDF Code Artifact (Available, Functional)

(2020). Finding the Right Cloud Configuration for Analytics Clusters. SoCC'20.

PDF

(2020). P4xos: Consensus as a Network Service. IEEE/ACM Transactions on Networking, 28(4).

PDF DOI

(2020). Do the Best Cloud Configurations Grow on Trees? An Experimental Evaluation of Black Box Algorithms for Optimizing Cloud Workloads. Proceedings of the VLDB Endowment, 13(11).

PDF Code Video DOI

(2020). Towards Consistent SDNs: A Case for Network State Fuzzing. IEEE Transactions on Network and Service Management, 17(2).

PDF DOI

(2020). A Survey on the Current Internet Interconnection Practices. ACM SIGCOMM Computer Communication Review, 50(1).

PDF DOI

(2020). On the Discrepancy between the Theoretical Analysis and Practical Implementations of Compressed Communication for Distributed Deep Learning. AAAI'20.

PDF Technical Report

(2019). IntML: Natural Compression for Distributed Deep Learning. SOSP AI Systems Workshop'19.

PDF

(2019). Cracking Open the Black Box: What Observations Can Tell Us About Reinforcement Learning Agents. NetAI'19.

PDF

(2019). Measurements as First-class Artifacts. INFOCOM'19.

PDF

(2018). Dynam-IX: a Dynamic Interconnection eXchange. CoNEXT'18.

PDF

(2018). Fast and Accurate Load Balancing for Geo-Distributed Storage Systems. SoCC'18.

PDF

(2018). Sonata: Query-Driven Streaming Network Telemetry. SIGCOMM'18.

PDF

(2018). Prelude: Ensuring Inter-Domain Loop-Freedom in SDN-Enabled Networks. APNet'18.

PDF

(2018). Dynam-IX: a Dynamic Interconnection eXchange. SIGCOMM'18.

PDF

(2018). Renaissance: Self-Stabilizing Distributed SDN Control Plane. ICDCS'18.

PDF

(2018). Picking a Partner: A Fair Blockchain Based Scoring Protocol for Autonomous Systems. ANRW'18.

PDF

(2018). Mitigating Network Side Channel Leakage for Stream Processing Systems in Trusted Execution Environments. DEBS'18.

PDF

(2018). Moving Bits with a Fleet of Shared Virtual Routers. IFIP Networking'18.

PDF

(2018). Methodology, Measurement and Analysis of Flow Table Update Characteristics in Hardware OpenFlow Switches. Computer Networks, 136.

PDF DOI

(2017). SIXPACK: Securing Internet eXchange Points Against Curious onlooKers. CoNEXT'17.

PDF

(2017). In-Network Computation is a Dumb Idea Whose Time Has Come. HotNets'17.

PDF

(2017). ENDEAVOUR: A Scalable SDN Architecture for Real-World IXPs. IEEE Journal on Selected Areas in Communications, 35(11).

PDF DOI

(2017). Kurma: Geo-Distributed Load Balancer for Back-End Storage Systems. SOSP'17.

PDF

(2017). Towards Automatic Parameter Tuning of Stream Processing Systems. SoCC'17.

PDF

(2017). Distributed Resource Management Across Process Boundaries. SoCC'17.

PDF

(2017). DAIET: A System for Data Aggregation Inside the Network. SoCC'17.

PDF

(2017). Privacy-Preserving Detection of Inter-Domain SDN Rules Overlaps. SIGCOMM'17.

PDF

(2017). Learning Reproducibility with a Yearly Networking Contest. Reproducibility'17.

PDF

(2017). A Self-Organizing Distributed and In-Band SDN Control Plane. ICDCS'17.

PDF

(2017). Towards meticulous data plane monitoring. EuroDW'17.

PDF

(2017). Rein: Taming Tail Latency in Key-Value Stores via Multiget Scheduling. EuroSys'17.

PDF

(2017). Decentralized Consistent Updates in SDN. SOSR'17.

PDF Technical Report

(2017). Correct by Construction Networks using Stepwise Refinement. NSDI'17.

PDF

(2016). Network Monitoring as a Streaming Analytics Problem. HotNets'16.

PDF

(2016). Inter-domain Networking Innovation on Steroids: Empowering IXPs with SDN Capabilities. IEEE Communications Magazine, 54(10).

PDF DOI

(2016). Towards Securing Internet eXchange Points Against Curious onlooKers. ANRW'16.

PDF

(2016). Towards Decentralized Fast Consistent Updates. ANRW'16.

PDF

(2016). Ground Control to Major Faults: Towards a Fault Tolerant and Adaptive SDN Control Network. DISN'16.

PDF

(2016). Network Hardware-Accelerated Consensus. arXiv:1605.05619.

PDF

(2016). Scheduling Multi-flow Network Updates in Software-Defined NFV Systems. SWFAN'16.

PDF

(2016). Paxos Made Switch-y. ACM SIGCOMM Computer Communication Review, 46(2).

PDF DOI Technical Report

(2016). Automagic Configuration Optimization for Big Data Processing Systems. EuroDW'16.

PDF

(2016). An Industrial-Scale Software Defined Internet Exchange Point. NSDI'16.

PDF

(2015). Systematically Testing OpenFlow Controller Applications. Computer Networks, 92(2).

PDF DOI

(2015). BRB: BetteR Batch Scheduling to Reduce Tail Latencies in Cloud Data Stores. SIGCOMM'15.

PDF

(2015). NetPaxos: Consensus at Network Speed. SOSR'15.

PDF

(2015). Medieval: Towards A Self-Stabilizing, Plug & Play, In-Band SDN Control Network. SOSR'15.

PDF

(2015). C3: Cutting Tail Latency in Cloud Data Stores via Adaptive Replica Selection. NSDI'15.

PDF

(2015). A Distributed and Robust SDN Control Plane for Transactional Network Updates. INFOCOM'15.

PDF Technical Report

(2014). Software-Defined Networks: Incremental Deployment with Panopticon. Computer, 47(11).

PDF DOI

(2014). SDLoad: An Extensible Framework for SDN Workload Generation. HotSDN'14.

PDF

(2014). ESPRES: Transparent SDN Update Scheduling. HotSDN'14.

PDF

(2014). Panopticon: Reaping the Benefits of Incremental SDN Deployment in Enterprise Networks. USENIX-ATC'14.

PDF Technical Report

(2014). STN: A Robust and Distributed SDN Control Plane. ONS'14.

PDF

(2014). ESPRES: Easy Scheduling and Prioritization for SDN. ONS'14.

PDF

(2013). Brief Announcement: Towards Distributed and Reliable Software Defined Networking. DISC'13.

PDF

(2013). Software Transactional Networking: Concurrent and Consistent Policy Composition. HotSDN'13.

PDF

(2013). OF.CPP: Consistent Packet Processing for OpenFlow. HotSDN'13.

PDF

(2013). Incremental SDN Deployment in Enterprise Networks. SIGCOMM'13.

PDF

(2013). FatTire: Declarative Fault Tolerance for Software-Defined Networks. HotSDN'13.

PDF

(2013). Automatic Failure Recovery for Software-Defined Networks. HotSDN'13.

PDF

(2013). Toward Transitional SDN Deployment in Enterprise Networks. ONS'13.

PDF

(2013). Systematic Software Testing Meets Networking. ONS'13.

PDF

(2013). Consistent Packet Processing — Because Consistent Updates Are Not Enough. ONS'13.

PDF

(2012). A SOFT Way for OpenFlow Switch Interoperability Testing. CoNEXT'12.

PDF

(2012). OFTEN Testing OpenFlow Networks. EWSDN'12.

PDF

(2012). A NICE Way to Test OpenFlow Applications. NSDI'12.

PDF Technical Report

(2011). Is Your OpenFlow Application Correct?. CoNEXT Student Workshop'11.

PDF

(2011). Identifying and Using Energy-Critical Paths. CoNEXT'11.

PDF

(2011). Finding Almost-Invariants in Distributed Systems. SRDS'11.

PDF

(2011). Automating the Testing of OpenFlow Applications. WRiPE'11.

PDF

(2011). Online Testing of Federated and Heterogeneous Distributed Systems. SIGCOMM'11.

PDF

(2011). Insomnia in the Access or How to Curb Access Network Related Energy Consumption. SIGCOMM'11.

PDF

(2011). Evaluation and Design of Cache Replacement Policies under Flooding Attacks. TRAC'11.

PDF

(2011). Toward Online Testing of Federated and Heterogeneous Distributed Systems. USENIX-ATC'11.

PDF Technical Report

(2011). Evolution of Cache Replacement Policies to Track Heavy-hitter Flows. PAM'11.

PDF

(2010). Evolution of Cache Replacement Policies to Track Heavy-hitter Flows. ANCS'10.

PDF

(2010). DiCE: Predicting Faults in Heterogeneous, Federated Distributed Systems. OSDI'10.

(2010). Fault Prediction in Distributed Systems Gone Wild. LADIS'10.

PDF

(2009). Experience with High-Speed Automated Application-Identification for Network-Management. ANCS'09.

PDF

(2009). Tracking Elephant Flows in Internet Backbone Traffic with an FPGA-based Cache. FPL'09.

PDF

(2009). Experiences with the Collection of Application Ground Truth Data. In RECIPE: Robust and Efficient traffic Classification of IP nEtworks.

(2009). GTVS: Boosting the Collection of Application Traffic Ground Truth. TMA'09.

PDF

(2009). Toward the Identification of Anonymous Web Proxies. PAM Student Workshop'09.

PDF

(2009). Efficient Application Identification and the Temporal and Spatial Stability of Classification Schema. Computer Networks, 53(6).

PDF DOI

(2009). Per Flow Packet Sampling for High-Speed Network Monitoring. COMSNETS'09.

PDF

(2008). Classifying HTTP Traffic in the New Age. SIGCOMM'08.

PDF

(2008). On the Double-Faced Nature of P2P Traffic. PDP'08.

PDF

(2008). Characterizing the network behavior of P2P traffic. IT-NEWS'08.

PDF

(2005). A High Performance IP Traffic Generation Tool Based on the Intel IXP2400 Network Processor. TIWDC'05.

PDF DOI

(1900). Application Aware Placement and Scheduling for Multi-tenant Clouds. HPI Future SOC Lab Technical Report.

PDF

Certain contents is © IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Certain contents is © ACM. This is the author’s version of the work. It is posted here for your personal use. Not for redistribution.