I've had the pleasure of working with great students.
Most of what I really understand in my field, I learned in working with them.

Bob Carter PhD 1998 Performance Measurement and Prediction in Packet-Switched Networks: Techniques and Applications Bob's thesis was focused on how to choose the best server when data is replicated in the Internet, which we called the “server selection problem.” To address this, Bob built, validated, and distributed the first robust tools for bottleneck bandwidth measurement. Subsequent to Bob's work, a number of other reseachers picked up the problem and it has become an active area of ongoing investigation. Bob is currently (2018) Software Engineering Manager at Leidos.

Paul Barford PhD 2000 Modeling, Measurement, and Performance of World Wide Web Transactions Paul's thesis was motivated by the simple question: “Why is the Web so slow?” - which turned out to have very interesting answers. In the process, Paul developed statistical models of Web workloads, a distributed Web measurement platform, and an innovative method for analyzing Web transactions based on the idea of critical path analysis. Paul is currently (2018) Professor, Department of Computer Science, University of Wisconsin (Madison).

Jun Liu PhD 2002 Characterizing Network Elements and Paths Using Packet Loss Behavior Jun's thesis work concentrated on trying to understand the conditions during packet loss in the Internet. He developed methods for estimating buffer sizes in routers, and for isolating lossses in wireless settings, based on the technique he developed called “loss pairs.” Jun is currently (2011) Associate Professor, Department of Computer Science, University of North Dakota.

Anukool Lakhina PhD 2006 Network-Wide Traffic Analysis: Methods and Applications Anukool's thesis research was concerned with a new kind of traffic analysis, namely network-wide. He developed the first methods for characterizing and understanding the ensemble of traffic flows in a network as a whole. He showed that the resulting methods have considerable power for detecting and classifying network anomalies. Anukool is currently (2018) Founder and President at Guavus, a Thales Company.

Nahur Fonseca PhD 2008 Stochastic Modeling Applied to Detection Problems in Network Protocols and Traffic Nahur's thesis research looked at ways that stochastic models could be used to improve network performance. He used Bayes detectors to improve TCP's packet loss identification algorithm, and he showed the existence of a new kind of long-range memory in network traffic, which has implications for anomaly detection. Nahur is currently (2018) Senior Performance Engineer at Akamai, Inc.

Vijay Erramilli PhD 2008 Forwarding in Mobile Opportunistic Networks Vijay's thesis research looked at a new kind of network: one formed by people as they carry around mobile devices like phones. He characterized these networks’ properties, and developed new, efficient forwarding algorithms for use in such networks. Vijay is currently (2023) Principal Data Scientist at Salesforce, Inc.

Gonca Gürsun PhD 2013 Inferring Hidden Features in the Internet Gonca's thesis research developed new ways of estimating properties of the Internet that are hard or impossible to directly measure. She showed how to infer both the path that traffic takes and the amount of traffic that flows in situations where those quantities can't be directly observed. To do this she developed new metrics for analyzing routing, and new methods of traffic inference. Gonca is currently (2023) Product Owner / Activity Lead at Bosch Center for Artificial Intelligence.

Andrej Cvetkovski PhD 2013 Graph Embeddings for Low-Stretch Greedy Routing Andrej's thesis research developed algorithms and heuristics to improve greedy routing – routing in which a packet is forwarded to the neighbor closest to the destination. Andrej's work made heavy use of hyperbolic geometry, and relied on embedding network nodes in the Poincare disk. He showed how to enable growing graphs, embed weighted graphs with low distortion, and achieve greedy routing with low stretch (short paths). Andrej is currently (2020) Associate Professor, Faculty of Computer Science, Mother Teresa University, Skopje, North Macedonia.

Natali Ruchansky PhD 2016 coadvised with Evimaria Terzi Matrix Completion with Structure Natali's thesis developed new insights into problems involving low-rank matrix completion. While most matrix completion work is statistical in nature, Natali took fundamentally combinatorial approaches, developing an organizing framework and then a new combinatorial algorithm for active matrix-completion. She also developed new methods for identifying low-rank submatrices in a partially-observed matrix. Natali is currently (2023) Manager, Machine Learning at Netflix.

Giovanni Comarela PhD 2017 On the Dynamics of Interdomain Routing in the Internet Giovanni's thesis work developed new tools (algorithms and metrics) to analyze the behavior of interdomain routing over time. He developed a data mining approach to detecting large-scale changes in Internet routing that re-occur many times over long timescales, and he showed how to identify the link or Autonomous System (AS) most likely to be reponsible for those events. He also developed new metrics and algorithms for indentifying unusually-routed ASes, and used them to shed light on how global routing patterns have evolved over a thirteen-year period. Giovanni is currently (2023) Assistant Professor of Computer Science at the Universidade Federal do Espirito Santo, Brazil.

Larissa Spinelli PhD 2019 Empirical Studies of Factors Affecting Opinion Dynamics Larissa's thesis asked about what happens when humans and Internet systems are connected in a feedback loop. She looked at how recommender systems and human opinions can each shape the other, how the features of online product reviews affect the evolution of ratings, and how recommender systems like YouTube guide users away from reliable information. Larissa is currently (2020) Software Engineer in Algorithms and Data Science at Wayfair.

Bashir Rastegarpanah PhD 2021 Tools for Responsible Decision-Making in Machine Learning Bashir's thesis developed new methods for improving the social and ethical impacts of machine learning systems. He developed a novel method for improving the fairness of recommender systems based on augmenting the training data, and a new formalization and associated auditing algorithms for verifying compliance of machine learning systems with the principle of data minimization. Bashir is currently (2023) Data Scientist at Fiddler AI.

Ahmed Youssef PhD 2023 coadvised with Andrew Emili Computational Models to Uncover Cell State Proteomes and Profile Protein Interaction Dynamics Ahmed's thesis developed a new analysis pipeline for studying the remodeling of protein interactions from dynamic CF/MS data. He also developed a novel deconvolution algorithm allowing bulk proteomics measurements paired with single-cell RNA profiles, to be separated into protein abundance measures on a cell-state basis. Ahmed is currently (2023) Bioinformatics Scientist at Acrivon Therapeutics.

If you'd like to read more about any of these projects,
all of the papers I have written with my students are available from my publications page.