Spotlights

Each month the center tells a story about one of its doctoral students, and often one of its faculty: what they built, why it matters, and where to read the work. Faculty nominate students by writing to the director.

Faculty spotlight, September 2026

Getting machines to agree, even when things go wrong

Lewis Tseng works on how groups of computers, from drone teams to blockchains, keep agreeing with each other when some of them fail or are attacked. In August, NSF funded his plan for a shared data system for microscopy that its users will help design.

Lewis Tseng
Lewis TsengAssociate Professor, Electrical and Computer Engineering

Electrical and Computer Engineering; co-lead, next-generation networks and trusted devices thrust

Lewis Tseng, Associate Professor of Electrical and Computer Engineering, studies a question that comes up wherever one job is split across many computers: how do they agree on what happened, and keep agreeing when some of them crash, fall behind, or are attacked? He came to UMass Lowell in 2024 after positions at Clark University, Boston College, and Toyota InfoTechnology Center. He co-leads the center's next-generation networks and trusted devices thrust with Orlando Arias, and this year he received the ECE department's Teaching Excellence Award.

The question

Drones searching an area together, edge computers running a control loop, and a blockchain recording transactions all depend on many machines staying consistent. If they stop agreeing, nothing built on top of them can be trusted. His NSF CAREER award, which began in 2024, looks at the basic coordination steps these systems rely on, and at how to keep them correct, and fast enough for real-time control, while someone is attacking the system.

Recent work

In the past two years his group has published on multi-drone search coordinated through satellites and edge computers (IEEE GLOBECOM 2025), digital twins run as a service in hybrid clouds (IEEE ICHMS 2025), intrusion detection that keeps working when some detectors fail (ICSC 2025), and ways to replicate data reliably across machines (USENIX NSDI 2025; ACM SoCC 2025). In August he presented work at an ACM SIGCOMM workshop on timing control in quantum cloud systems, where entanglement decays if it isn't used in time.

New in August: a data system its users help design

In August, NSF's Office of Advanced Cyberinfrastructure awarded Tseng $499,490 over two years to plan a federated, AI-ready data system for advanced microscopy and imaging. His co-PIs are Hsien-Yuan (Mark) Hsu of the School of Education and Yu-Tsun Shao of the University of Southern California.

The plan is for data to stay near the instruments and institutions that produce it, with shared services that help researchers find it, trace where it came from, trust it, and reuse it with AI tools. The team will design it with the scientists who would use it: first helping them picture what such a system could do, then studying how they work and what worries them, then designing the architecture, operating model, and responsible AI practices together. Each lab keeps control of its own data, which makes it the same kind of trust problem Tseng works on in distributed systems.

Engineering and education in one group

His group spans two colleges. Kritee Neupane, a computer engineering Ph.D. student, works on networked digital twins, blockchain provenance, and fault tolerance. Layann Shaban began his Ph.D. in 2026 after co-authoring the IEEE GLOBECOM drone paper. Negin Yazdani Motlagh and Youlim Lee are Education Ph.D. students co-advised with Hsu: Negin studies AI and computational thinking in STEM education, and Youlim studies AI literacy and the ethical use of AI in research.

Why it matters to the center

Grid controllers, autonomous robots, and the networks between them all depend on devices agreeing on what is true, sometimes while an attacker tries to split them. That is the problem Tseng works on, and it is why he co-leads the trusted devices thrust and is a Co-PI on SUMMIT, the center's three-university grid testbed.

The papers

Student spotlight, September 2026

What employers ask for, in their own words

Negin Yazdani Motlagh went through 82,624 job posts from 2012 to 2025 to see which soft skills computing employers ask for, and how often.

Negin Yazdani Motlagh
Negin Yazdani MotlaghPh.D. student, Research and Evaluation in Education, joined 2024; co-advised by Hsien-Yuan (Mark) Hsu

Advised by Hsien-Yuan (Mark) Hsu and Lewis Tseng, School of Education and Electrical and Computer Engineering

Engineering programs are asked to teach communication, teamwork, and problem solving alongside technical skills. Negin Yazdani Motlagh, a Ph.D. student in Research and Evaluation in Education since 2024, wanted to know which of those skills employers mention when they hire. She is co-advised by Mark Hsu of the School of Education and Lewis Tseng of Electrical and Computer Engineering, and her research focuses on artificial intelligence and computational thinking in STEM education.

The data

She used the monthly "Who Is Hiring" threads on Hacker News, run by the startup accelerator Y Combinator, where companies share short job posts. Because the posts are short, what an employer includes says something about what it cares about. The study covers 82,624 computing and computer engineering postings from 2012 to 2025.

The method

Each post was checked against a dictionary of soft skills sorted into three groups: Interpersonal & leadership, Personal effectiveness & growth, and Conceptual/thinking. Two measures follow each group over time: coverage, the share of posts that mention it, and intensity, the average number of mentions per post.

What she found

Interpersonal and leadership skills came first on both measures, for every kind of job and in every year, with personal effectiveness and growth second. Conceptual and thinking skills showed up less often, but they never dropped out, and they were more common in AI jobs. The paper offers these trends as a reference for instructors deciding how their computing and computer engineering courses handle professional skills.

How the work grew

The study started as a work-in-progress paper at the 2025 IEEE Frontiers in Education Conference, covering 2012 to 2024, written with Kritee Neupane, a computer engineering Ph.D. student in Tseng's group. The full study, with Hsu and Tseng, was published in September in the Computers in Education Journal (CoED) of the American Society for Engineering Education. She also co-wrote a 2026 ACM Sustainability Week paper with Tseng on the energy cost of how blockchains reach agreement.

Why it matters to the center

Hsu leads the center's education and workforce work. Negin's study gives that work a measured starting point: what employers actually write when they hire.

The papers

Student spotlight, September 2026

Building the networks behind AI

Three journal articles in six months on how fiber-optic networks can keep up with AI, and an open-source platform the field can build on.

Arash Rezaee
Arash RezaeePh.D. Candidate, joined 2022
Ryan McCann
Ryan McCannPh.D. Candidate, joined 2024

Advised by Vinod M. Vokkarane, Advanced Communication Networks Laboratory

Every AI application, cloud service, and connected device depends on something most people never see: the communication networks carrying enormous amounts of data behind the scenes. Two doctoral researchers in the center's Advanced Communication Networks Laboratory, Arash Rezaee and Ryan McCann, have spent the past year working on how those networks keep up.

Their work has produced three articles accepted in the last six months by the Journal of Optical Communications and Networking, published jointly by IEEE and Optica. Arash is first author on all three. Ryan co-led the work on FUSION, the group's open-source optical networking platform, and contributed to the other two studies. Collaborators at the University of Virginia and in Spain joined the work. Both students are advised by Vinod Vokkarane, the center's director.

The question

Future AI systems will need more than powerful processors and large data centers. They will need networks that move huge volumes of information quickly and reliably between users, edge devices, cloud platforms, and computing facilities. There are two main ways to add capacity to a fiber network: use a wider range of the optical spectrum by transmitting across several wavelength bands, or add spatial capacity by sending data through multiple cores inside one fiber or through parallel fiber pairs.

Both work. Combining them is hard. Signals interfere with one another, performance varies across the spectrum, and some of the added capacity turns out to be unusable. The three papers take on that problem from three directions: long-term planning, real-time operation, and reproducible research.

Planning: capacity on paper is not capacity in practice

The first article develops a planning framework that evaluates optical bands and spatial channels together rather than as separate upgrades. It accounts for signal noise, nonlinear effects, power changes, equipment limits, and interference between neighboring fiber cores, and parts of it were validated on a commercial multicore fiber. Its central finding is that adding bands or cores does not always add usable capacity; the best choice depends on the spectrum in use, the structure of the fiber, and the physical conditions in the network. The article is an invited extension of the team's paper at ECOC 2025.

Operation: smarter decisions while the network runs

The second article turns to the decisions a network makes as traffic arrives. A quality-of-transmission-aware controller chooses the route, band, spatial lane, frequency channel, and modulation format for each connection, grooms new traffic into unused capacity on existing connections, and slices large requests across channels when no single channel can carry them. Comparing multi-core fibers with parallel fiber pairs, the paper shows the conditions under which each performs best rather than declaring a winner. It extends the team's paper at ONDM 2025.

Research: a shared platform, FUSION

The third article introduces FUSION, the Flexible Unified Simulator for Intelligent Optical Networking, which Ryan co-founded and leads as its principal developer. Optical networking studies are hard to compare because groups use different software, assumptions, and network configurations. FUSION gives researchers one modular environment covering conventional, multi-band, and space-division multiplexed networks, with physical-layer models, control algorithms, grooming, survivability, and AI-based decision making, plus automated testing and deterministic replay so experiments can be checked and repeated. It was the platform behind the other two papers, and it is released as open source for others to extend. The article appears in JOCN's feature issue on benchmarking in optical networks.

Why it matters to the center

Smart grids, autonomous platforms, connected health care, and intelligent transportation all depend on networks that move data reliably and respond when conditions change. Arash and Ryan's work connects fiber physics to network control and to software the field can share. It is the communications foundation that the center's cyber-physical systems stand on.

The papers

UMass Lowell's contributions were supported in part by the National Science Foundation under Award No. 2008530.