Emon Dey

Postdoctoral Appointee, Argonne National Laboratory

Emon Dey

AI researcher specializing in large-scale distributed training, scientific machine learning, federated learning, robotics, and efficient deep learning. I build robust ML systems for scientific imaging, autonomous systems, power-grid analytics, and cross-facility distributed computing.

Email Google Scholar GitHub LinkedIn Resume

Updates

Recent News

About

I am a Postdoctoral Appointee in the Mathematics and Computer Science Division at Argonne National Laboratory. I received my Ph.D. in Information Systems from the University of Maryland, Baltimore County, where I worked with the Center for Real-time Distributed Sensing and Autonomy and the Mobile, Pervasive, and Sensor Computing Lab advised by Dr. Nirmalya Roy.

My work focuses on scalable AI training architectures for leadership-class HPC systems, scientific imaging, smart power-grid analytics, queue-aware federated learning, and robust communication protocols for distributed autonomous systems.

During my Ph.D., I developed robust and communication-efficient federated learning methods for non-IID data, lossy communication channels, model compression, adaptive quantization, and ROS/ROS2-based heterogeneous robot experimentation. I received my M.S. from UMBC in 2022 and my B.Sc. in Electrical and Electronic Engineering from Chittagong University of Engineering and Technology.

A recurring part of this work is making FL communication robust in contested and lossy environments, including packet-loss scenarios caused by high velocity differences among collaborating robotic agents.

High Performance Computing Federated Learning Deep Model Compression Robotics Generative AI

Scientific machine learning

Multi-modal and physics-aware Vision Transformers for ptychographic phase retrieval and real-time diffraction imaging.

Federated learning

Queue-aware asynchronous FL, continual FL, class-aware scaling, and federated training for smart-grid systems.

Robotics and CPS

Robust learning and ROS/ROS2 middleware for heterogeneous UAV/UGV experimentation.

Education

  • Ph.D. in Information Systems, AI/ML Track University of Maryland, Baltimore County, Apr. 2025

    Dissertation: A Robust and Resilient Federated Learning Framework over Lossy Communication Channels.

    Advisor: Dr. Nirmalya Roy.

  • M.S. in Information Systems, AI/ML Track University of Maryland, Baltimore County, May 2022
  • B.Sc. in Electrical and Electronic Engineering Chittagong University of Engineering and Technology, Oct. 2017

All Publications

  • ICML 2026: Li, Y., Dey, E., Li, Z., Raghavan, K., Madduri, R., and Kim, K. "FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training."

  • PerCom 2026: Dey, E., Ravi, A., Shinde, G., Chugh, G., Ghosh, I., Misra, A., and Roy, N. "Fed-CASQ: Enhancing Class-Wise Accuracy in Pervasive Federated Learning with Class-Aware Scaling and Quantization."

  • ISVLSI 2026: Dey, E., Hasan, Z., and Roy, N. "FedGALP: Federated Generative Adversarial Learning for Remote Photoplethysmography."

  • ICCCN 2026: Hossain, J., Dey, E., Chugh, S., Ahmed, M., Anwar, M. S., Faridee, A. Z., and Roy, N. "SERN: Simulation-Enhanced Realistic Navigation for Multi-Agent Robotic Systems in Contested Environments."

  • WIREs DMKD 2025: Shinde, G., Ravi, A., Dey, E., Sakib, S., Rampure, M., and Roy, N. "A Survey on Efficient Vision-Language Models."

  • ACM TAAS 2025: Anwar, M. S., Ravi, A., Dey, E., Ghosh, I., Ahmed, M., and Roy, N. "MobHeteroCAS: Mobility-Aware DNN Task Scheduling in Heterogeneous Collaborative Autonomous Systems."

  • PeRConAI 2025: Shinde, G., Ravi, A., Dey, E., Lewis, J., and Roy, N. "TAVIC-DAS: Task and Channel-Aware Variable-Rate Image Compression for Distributed Autonomous System."

  • SMARTCOMP 2025: Sakib, S., Shinde, G., Dey, E., and Roy, N. "E2RespUNet: End-to-End Respiratory Signal Reconstruction and Rate Prediction Using a Unified Attention-Enhanced U-Net."

  • DCOSS-IoT 2025: Anwar, M. S., Ravi, A., Dey, E., Ghosh, I., Ahmed, M., and Roy, N. "CoOpTex: Multimodal Cooperative Perception and Task Execution in Time-critical Distributed Autonomous Systems."

  • COMSNETS 2025: Dey, E., Ravi, A., Lewis, J., Kumar, V. K., Freeman, J., Gregory, T., Suri, N., Busart, C., and Roy, N. "DACC-Comm: DNN-Powered Adaptive Compression and Flow Control for Robust Communication in Network-Constrained Environments."

  • ICCCN 2023: Dey, E., Walczak, M., Anwar, M. S., Roy, N., Freeman, J., Gregory, T., and Busart, C. "A Novel ROS2 QoS Policy-Enabled Synchronizing Middleware for Co-Simulation of Heterogeneous Multi-Robot Systems."

  • MASS 2023: Anwar, M. S., Dey, E., Devnath, M. K., Ghosh, I., Khan, N., Freeman, J., Gregory, T., Suri, N., Jayaraja, K., Ramamurthy, S. R., and Roy, N. "HeteroEdge: Addressing Asymmetry in Heterogeneous Collaborative Autonomous Systems."

  • SMARTCOMP 2023: Devnath, M. K., Chakma, A., Conn, M., Hasan, Z., Dey, E., Anwar, M. S., Pal, B., and Roy, N. "A Systematic Study on Object Recognition Using Millimeter-wave Radar."

  • CVPRW 2023: Ovi, P. R., Dey, E., Roy, N., and Gangopadhyay, A. "Mixed Quantization Enabled Federated Learning to Tackle Gradient Inversion Attacks."

  • SPIE 2022: Ovi, P. R., Dey, E., Roy, N., Gangopadhyay, A., and Erbacher, R. F. "Towards Developing a Data Security Aware Federated Training Framework in Multi-modal Contested Environments." Best Paper Award.

  • DCOSS 2022: Dey, E., Hossain, J., Roy, N., and Busart, C. "SynchroSim: An Integrated Co-simulation Middleware for Heterogeneous Multi-robot System."

  • SMARTCOMP 2022: Hasan, Z., Dey, E., Ramamurthy, S. R., Roy, N., and Misra, A. "RhythmEdge: Enabling Contactless Heart Rate Estimation on the Edge." Best Paper Award.

  • SMARTCOMP 2022 Demo: Hasan, Z., Dey, E., Ramamurthy, S. R., Roy, N., and Misra, A. "Demo: RhythmEdge: Enabling Contactless Heart Rate Estimation on the Edge."

  • SMARTCOMP 2021: Ovi, P. R., Dey, E., Roy, N., and Gangopadhyay, A. "ARIS: A Real-Time Edge Computed Accident Risk Inference System."

  • BIBE 2020: Dey, E., and Roy, N. "OMAD: On-device Mental Anomaly Detection for Substance and Non-Substance Users."

  • I2CT 2019: Mazumder, A. N., Dey, E., and Majumder, S. "A Color Image Watermarking Scheme Employing the Features of Directive Contrast in the DWT-SVD Domain."

  • EICT 2017: Dey, E., Majumder, S., and Mazumder, A. N. "A New Approach to Color Image Watermarking Based on Joint DWT-SVD Domain in YIQ Color Space."

  • Under Review: Raghavan, K., Dey, E., Hahn, S., Munson, T., Li, Y., and Kim, K. "The Foundations of Ensemble Continual Learning in Federated Learning."

Technical Projects

Project 1: SYNAPS-I: Synergistic Neutron and Photon Autonomous Science-Imaging Oct. 2025 - Ongoing
  • Designed a physics-aware Vision Transformer for 2D ptychography generalized across thousands of probes.
  • Optimized the ViT model for large-scale training on Perlmutter and Aurora, scaling up to 24,576 GPUs.
  • Designed a fine-tuning pipeline with the trained ViT model to facilitate real-time deployment at the beamline.
Media Code
Project 2: GridMind: Powering the Control Room of the Future with AI Agents Mar. 2026 - Ongoing
  • Investigating data and model scaling performance of the proposed Lumina model with heterogeneous graph architecture.
  • Designing IsoFLOP experiments to reveal compute-optimal balance between model size and training data allocation.
  • Deploying the Lumina model across multiple DOE supercomputers to conduct federated training.
Media Code
Project 3: AI4S Continual Federated Learning under Intermittent Client Participation Jun. 2025 - Ongoing
  • Collaborating to develop the foundations of ensemble continual learning in federated learning settings.
  • Explored the impact of arbitrary arrival and dropping-off of client nodes on server aggregation.
  • Designed the experimental setting to test the proposed ensemble continual learning algorithm.
Project page Code Related manuscript
Project 4: AI4S Asynchronous Federated Learning for High Performance Computing Aug. 2025 - Ongoing
  • Tackled varied queuing time across multiple DOE supercomputers while deploying federated learning.
  • Implemented a queue-delay-aware asynchronous training framework and tested it in simulation.
  • Successfully deployed the proposed framework across Aurora, Perlmutter, Polaris, and Frontier.
Project page Code Related publication
Remote Robotic Experimentation using Distributed Virtual Proving Ground, DEVCOM ARL Summer 2021 - Spring 2025
  • Demonstrated components that leverage ARL's Distributed Virtual Proving Ground infrastructure.
  • Conducted field experimentation to demonstrate class-incremental federated learning.
  • Deployed a lightweight simultaneous localization and mapping system on Clearpath Jackal and SPOT.
  • Monitored robotic agents' movements and actions from a long-distance remote location.
Project page Code Media
Robust Deep Learning Models for Securing Multi-domain Autonomous Cyber-Physical Systems, ONR Spring 2023 - Spring 2025
  • Developed a model-poisoning-attack-resistant federated learning approach for camouflaged and occluded object detection.
  • Developed a robust federated learning approach for camouflaged and occluded multi-class object detection.
  • Integrated adaptive quantization-based compressive modeling to reduce communication overhead.
  • Tackled global and local class imbalances using customized Layer-wise Relevance Propagation.
Project page Code Related publications
Resilient and Effective Communication and Sensing for Autonomous Platforms and Systems, NSF Spring 2021 - Spring 2025
  • Explored communication challenges in non-line-of-sight scenarios within ROS-compatible networks.
  • Designed data-security-aware low-power communication protocols in adversarial contexts.
  • Proposed a modified sliding-window protocol for high velocity disparity among robotic agents.
Project page Code Related publications
Asynchronous Federated Training in Multi-Modal Contested Environment, DEVCOM ARL Fall 2022 - Spring 2024
  • Developed a multi-modal deep model for object detection using image and acoustic data in contested environments.
  • Addressed data distribution, synchronization, communication overhead, and security issues for asynchronous federated learning.
  • Proposed a multi-modal federated training setup to reduce resource overhead using 2-bit quantization-based compression.
Project page Code Related publication

Project Showcase

  • August 1, 2024: Cross-institutional Remote Robotics Experimentation using ARL's Distributed Virtual Proving Ground Facility, ARL R2C2, Graces Quarters, Maryland.
  • August 2, 2023: Enhancing Situational Awareness with Federated Learning using DVPG, ARL R2C2, Graces Quarters, Maryland.
  • August 2, 2022: AI-Enabled Decision Making in Multiple Domains, ARL R2C2, Graces Quarters, Maryland.
  • April 27, 2022: Deploying Federated Learning via ARL's DVPG Network, RetriEVER Empowered, UMBC.

Experience

  • Postdoctoral Appointee, MCS Division Argonne National Laboratory, Apr. 2025 - Present, Lemont, IL

    Advisor: Dr. Kibaek Kim.

    Working on multi-modal Vision Transformers for diffraction imaging, queue-aware asynchronous federated learning for HPC, federated learning for electrical smart-grid systems, and continual federated learning.

  • Graduate Research Assistant, CARDS Lab University of Maryland, Baltimore County, Aug. 2021 - Apr. 2025, Baltimore, MD

    Conducted research on decentralized multi-master ROS systems for communication between UAVs and UGVs, remote robotic experimentation, and robust federated learning for autonomous systems.

  • Graduate Research Assistant, MPSC Lab University of Maryland, Baltimore County, Aug. 2019 - Jul. 2020, Baltimore, MD

    Worked on accelerating transfer learning on edge devices by reducing negative transfer and developed DeepIoT for human activity recognition on IoT devices using deep models.

  • Graduate Teaching Assistant, Department of Information Systems University of Maryland, Baltimore County, Aug. 2020 - May 2021, Baltimore, MD

    Supported courses on Decision Support Systems, Machine Learning Basics, and Database Application Development.

  • Lecturer Port City International University, Sep. 2018 - Jul. 2019

    Taught undergraduate electrical engineering courses including machine design, power systems, and digital signal processing.

  • Student Mentor, NSF-REU University of Maryland, Baltimore County, Summer 2021, 2022, 2023

    Supervised student projects in multi-robot middleware, contactless heart-rate monitoring, quantized federated learning, and statistical heterogeneity in federated learning for sea-ship datasets.

Teaching and Mentoring

Lecturer, Port City International University

  • EEE 463: High Voltage Engineering
  • EEE 316: Machine Design
  • EEE 320: Electrical Services Design
  • EEE 465: Power System II
  • BUS 217: Professional Ethics
  • EEE 325: Digital Signal Processing I

Teaching Assistant, UMBC

  • IS425: Decision Support Systems
  • Machine Learning Basics
  • Database Application Development

NSF-REU Mentor, UMBC

  • Synchronizing middleware for heterogeneous multi-robot systems with low-latency co-simulation.
  • Federated learning for camera-based contactless heart-rate monitoring.
  • Resource-efficient quantized federated learning for brain-tumor segmentation.
  • Addressing statistical heterogeneity in federated learning for sea-ship datasets.

Technical Skills

ProgrammingPython, C, C++, MATLAB, R, Bash, LaTeX
High Performance ComputingMPI, OpenMP, PyTorch Distributed, Slurm, PBS, NCCL/CCL, multi-node GPU/XPU training
AI/MLPyTorch, TensorFlow, Keras, Vision Transformers, Graph Neural Networks, Federated Learning, Generative AI, model compression, quantization
SimulationGazebo, MAVROS, PX4, Unity3D, NS-3, WEKA, Tableau
Specialized ToolsOpenCV, TFF, PySyft, FedML, APPFL, TF-Lite, ONNX, Q-Keras, ROS/ROS2
Hardware and SystemsAurora, Polaris, Perlmutter, Frontier, Raspberry Pi, Jetson Nano, Google Coral Dev Board, Intel NCS2, Arduino
Engineering ToolsLinux, Docker, Git, shell scripting, Weights & Biases
Soft SkillsLeadership, communication, goal-oriented execution, problem solving, adaptability

Presentations and Talks

  • April 15, 2025: "Quantized Federated Learning over Data Imbalance and Lossy Communication Channels," IS 698/800 Advanced Deep Learning Techniques.
  • October 17, 2024: "Introduction to Crazyflie Functionalities and ROS Integration," IS 709/809 Computational Methods for IS Research.
  • September 17, 2024: "Towards Resilient and Secure Federated Learning: Tackling Non-IID Data, Communication Challenges, and Data Security Threats," Argonne MCS Candidate Seminar.
  • February 21-22, 2024: "Resilient and Effective Communication and Sensing for Autonomous Vehicles and Systems," ArtIAMAS Winter Meeting and Annual Review.
  • February 13-14, 2023: "A Reliable and Low-latency Synchronizing Middleware for Co-simulation of a Heterogeneous Multi-robot System," ArtIAMAS Year 3 Kickoff Meeting.
  • September 29, 2022: "Robot Operating System (ROS)," IS 709/809 Computational Methods for IS Research.
  • March 16-17, 2022: "Decentralized Multi-master ROS System for Communication Between UGVs and UAVs," ArtIAMAS Annual Review.
  • November 3, 2020: "Neural Networks," IS 425 Decision Support Systems.

Professional Activities

  • Reviewer for TMC, TIOT, TIST, Pervasive and Mobile Computing, ACM ICDCN, IEEE Intelligent Environments, BDCAT, IEEE PerCom, IEEE SMARTCOMP, DCOSS-IoT, COMSNETS, IEEE CHASE, REU Symposium, and WristSense workshop.
  • Student volunteer at IEEE PerCom 2025 in Washington, DC.
  • Student mentor for NSF-REU projects in federated learning, robotics middleware, and edge sensing at UMBC.
  • Student member of the Society of Photo-Optical Instrumentation Engineers.