Submitted GPU-OASIS and When EV Charging Helps (EcoCharge) to the ACM e-Energy 2027 Fall Cycle.
Assistant Professor of Computer Science · Texas Lutheran University
Zahidur Talukder
Trustworthy & Sustainable AI Systems
I design learning and computing systems that remain useful under real constraints on privacy, security, fairness, hardware, communication, energy, water, and carbon. My work spans federated learning, robust aggregation, systems measurement, and environmentally aware computing.
News
Recent updates
Led a TLU Faculty–Student Summer Research project on joint carbon- and stress-adjusted-water-aware EV charging.
Balancing Bits and Drops appeared at ACM e-Energy 2026.
FedSRC appeared at IEEE EDGE 2025, and FairHetero appeared in ACM TOMPECS.
Selected as an MLCommons Machine Learning & Systems Rising Star.
01 · Research
Two connected research directions.
A common question connects my work: how should AI adapt when data, devices, infrastructure, and environmental conditions are heterogeneous?
I
Trustworthy Federated & Distributed AI
Secure, privacy-preserving, fair, and robust learning under statistical and systems heterogeneity—including corrupted or poisoned client data, unreliable updates, and unequal device capability.
- Robust federated aggregation
- Client self-regulation and participation
- Hardware-sensitive fairness
- Privacy and security under heterogeneity
II
Sustainable AI Systems & Infrastructure
Resource-aware computing that jointly reasons about water stress, carbon, electricity cost, accelerators, workload flexibility, and operational feasibility.
- Stress-adjusted water accounting
- GPU and workload scheduling
- Carbon + water co-optimization
- Low-cost systems measurement
Bridge project
SAW-Fed
Ongoing work on environmentally aware federated learning where client selection balances learning utility, participation fairness, and environmental cost without moving private data.
02 · Selected Work
Selected research
A compact view of the projects that best represent my current research trajectory.
Balancing Bits and Drops
Stress-adjusted water accounting for data centers, separating on-site cooling from electricity-related water impacts and making geographic and temporal water stress part of operational decisions.
FairHetero
Hardware-sensitive fairness in heterogeneous federated learning when resource-constrained clients cannot all execute the same model architecture.
FedSRC
Self-adaptive federated learning that moves quality-aware control to clients so they can regulate local training and participation when local data quality is unfavorable.
FedASL
Computationally efficient auto-weighted aggregation that adapts server-side weighting to reduce the effect of unfavorable or corrupted client updates.
Low-Cost Server Power Monitoring
Server-level power monitoring using conducted electromagnetic interference, connecting sustainable systems optimization to practical measurement.
GPU-OASIS + EcoCharge
Current work on feasibility-aware geo-distributed AI scheduling and carbon/water-stress-aware EV charging.
e-Energy 2027 submissions
03 · Teaching
Teaching & mentoring
I emphasize explanation, implementation, evaluation, and the assumptions behind technical choices.
I want students to move from understanding a concept to being able to explain why it works, implement or evaluate it, and recognize when its assumptions fail. Research mentorship extends the same philosophy through progressively greater ownership.
Student research in STAIL ↗CSCI 379Artificial Intelligence
CSCI 436Computer & Network Security
CSCI 332Computer Networks
CSCI 248Object-Oriented Programming
STAT 374Statistics
CSCI 437Senior Seminar & Research Project
Sustainable Trustworthy AI Lab
STAIL
My lab connects robust and privacy-preserving federated learning with sustainable AI systems, systems measurement, and undergraduate research.
Visit the lab ↗04 · Funding
Research support
Faculty–Student Summer Research Grant
Texas Lutheran University · approximately $10,000 · joint carbon- and stress-adjusted-water-aware EV charging.
External proposal development
Stress-adjusted water sustainability for data centers and AI infrastructure.
NSF-supported research contributions
Water sustainability and conducted-EMI-based server power monitoring.
05 · Talks
Invited talks
Addressing Water Footprint for Growing AI
MLCommons Rising Stars · NVIDIA Headquarters
Data Quality and Hardware-Sensitive Fairness in Heterogeneous FL
Ph.D. Lightning Talk · The University of Texas at Arlington
Auto-Weighted Aggregation for Heterogeneous FL
Massachusetts General Hospital / Harvard Medical School
Contact