image

Muhammad Umer

PhD at Stanford University
mumer@stanford.edu   +1 (650) 444-3722

About Me

I am a PhD student in Electrical Engineering at Stanford University, advised by Prof. John Cioffi. My research explores intelligent communications, optimization and control, AI/ML, reinforcement learning, and algorithmic reasoning. I actively collaborate with industry partners including Intel Corporation, Samsung, Google DeepMind, and Ericsson to bridge frontier research with practical applications.

Before Stanford, I worked as an R&D engineer at Adept (ATS) Inc., where I developed tools for synthetic data generation. I hold a Bachelor's degree in Electrical Engineering from the National University of Sciences & Technology (NUST), Pakistan, where I was awarded the Rector's Gold Medal for outstanding undergraduate thesis and the Chancellor's Silver Medal for academic excellence.

Experience

Sep. 2025 - present
Machine Learning & Communications Lab (MLC)
Stanford University, USA

Research Assistant

  • Working on optimizations for minPMAC to realize practical deployments of generalized decision feedback equalizers (GDFE) with weighted energy minimization and rate constraints for multi-user MIMO systems.
  • Developing neural Gaussian radio fields (nGRF) as a real-time channel estimation framework that uses Gaussian primitives to reduce feedback overhead and inference latency.
  • Designing Transformer architectures for wireless channel prediction to forecast future channel responses in non-stationary environments and enable proactive link adaptation.
July 2024 - June 2025
Adept (ATS) Inc.
National Science and Technology Park, Pakistan

R&D Engineer

  • Contributed to developing Ascend, a proprietary tool for generating synthetic data while preserving the statistical properties of original datasets.
  • Implemented correlative masking techniques to maintain data integrity during the synthesis process.
  • Led the integration of generative AI capabilities into Ascend to create high-fidelity synthetic data from limited source data.
May 2023 - June 2024
Information Processing and Transmission Lab (IPT)
NUST, Pakistan

Research Assistant

  • Developed novel network architectures and evaluated their performance through simulations and analytical methods.
  • Addressed emerging challenges in next-generation wireless networks through targeted research initiatives.
  • Investigated the application of deep reinforcement learning for optimizing resource allocation in future mobile networks.
June 2022 - Mar. 2023
TUKL Research and Development Lab
RPTU Kaiserslautern, Germany

Research Assistant

  • Advanced table recognition capabilities in image-based documents to automate information extraction from complex layouts.
  • Developed a tabular image generative augmentation technique to train Transformer architectures effectively under data-limited conditions.
  • Designed a convolution-less vision Transformer architecture specifically for table recognition tasks.

Select Publications

  • On the Fundamental Limits of LLMs at Scale
    MA Mohsin, M Umer, A Bilal, et al.
    arXiv preprint arXiv:2511.12869, 2025
  • Conditional Prior-based Non-stationary Channel Estimation Using Accelerated Diffusion Models
    MA Mohsin, A Bilal, M Umer, A Aali, MA Jamshed, DF Hougen, and JM Cioffi
    arXiv preprint arXiv:2509.15182, 2025
  • Transformer-based Sparse CSI Estimation for Non-stationary Channels
    MA Mohsin, M Umer, A Bilal, H Rizwan, S Bhattacharya, MA Jamshed, and JM Cioffi
    arXiv preprint arXiv:2511.01333, 2025
  • Neural Gaussian Radio Fields for Channel Estimation
    M Umer, MA Mohsin, A Bilal, and JM Cioffi
    arXiv preprint arXiv:2508.11668, 2025
  • Meta-Thinking in LLMs via Multi-Agent Reinforcement Learning: A Survey
    A Bilal, MA Mohsin, M Umer, MAK Bangash, and MA Jamshed
    arXiv preprint arXiv:2504.14520, 2025
  • Hierarchical Deep Reinforcement Learning for Adaptive Resource Management in Integrated Terrestrial and Non-Terrestrial Networks
    MA Mohsin, H Rizwan, M Umer, S Bhattacharya, A Bilal, and JM Cioffi
    AAAI Workshop on AI for Wireless Communications and Networking (AI4WCN), 2025
  • LLM-based Retrieval-Augmented Generation: A Novel Framework for Resource Optimization in 6G and Beyond Wireless Networks
    HMA Zeeshan, M Umer, M Akbar, A Kaushik, MA Jamshed, H Jung, and SA Hassan
    IEEE Communications Magazine, 63(10), 60-67, 2025
  • Intelligent Spectrum Sharing in Integrated TN-NTNs: A Hierarchical Deep Reinforcement Learning Approach
    M Umer, MA Mohsin, AA Nasir, H Abou-Zeid, and SA Hassan
    IEEE Wireless Communications, 32(3), 64-71, 2025
  • Reconfigurable Intelligent Surface-Assisted Aerial Nonterrestrial Networks: An Intelligent Synergy With Deep Reinforcement Learning
    M Umer, MA Mohsin, A Kaushik, QUA Nadeem, AA Nasir, and SA Hassan
    IEEE Vehicular Technology Magazine, 2025
  • Continual Learning for Wireless Channel Prediction
    MA Mohsin, M Umer, A Bilal, MA Jamshed, and JM Cioffi
    ICML Workshop on Machine Learning for Wireless Communication and Networks (ML4Wireless), 2025
  • Transforming ISAC with STAR-RIS: Design, Challenges, and Opportunities
    M Umer, S Basharat, SA Hassan, A Mahmood, and M Gidlund
    IEEE Network, 2024
  • Analysis of STAR-RIS-Assisted Downlink CoMP-NOMA Multi-Cell Networks under Nakagami-m Fading
    M Umer, MA Mohsin, M Gidlund, H Jung, and SA Hassan
    IEEE Communications Letters, 28(5), 1009-1013, 2024
  • Deep Reinforcement Learning for Trajectory and Phase Shift Optimization of Aerial RIS in CoMP-NOMA Networks
    M Umer, MA Mohsin, A Mahmood, K Dev, H Jung, M Gidlund, and SA Hassan
    IEEE Global Communications Conference (GLOBECOM), 79-84, 2024
  • Performance Analysis of STAR-RIS Enhanced CoMP-NOMA Multi-Cell Networks
    M Umer, MA Mohsin, SA Hassan, H Jung, and H Pervaiz
    IEEE Global Communications Conference (GLOBECOM), 2000-2005, 2023
  • PyramidTabNet: Transformer-Based Table Recognition in Image-Based Documents
    M Umer, MA Mohsin, A Ul-Hasan, and F Shafait
    International Conference on Document Analysis and Recognition (ICDAR), 420-437, 2023

Teaching

Sep. 2024 - Jan. 2025
School of Electrical Engineering and Computer Science, NUST

Teaching Assistant

  • Mobile Communication Systems (Fall 2024)

Service

  • Reviewer. ICC '25, GLOBECOM '24, IEEE Vehicular Technology Conference, ICDAR '23, GLOBECOM '23, IEEE Transactions on Vehicular Technology
  • Mentor. Adept Summer Internship Program (2024), IPT Undergraduate Research Program (2024)
  • Organizer. STEM Workshop @ The Mashal School (2024)

Awards & Accolades

  • Recipient of IEEE Honorary Certificate of Appreciation (2025)
  • Recipient of Stanford EE Department Fellowship (2025)
  • Recipient of Rector's Gold Medal for outstanding undergraduate thesis (2024)
  • Recipient of Chancellor's Silver Medal for academic excellence (2024)
  • Recipient of Financial Award for Undergraduate Research (2024)
  • Recipient of NUST Merit Scholarship (2021 to 2024)
  • Recipient of Travel Award for KAIST EE Camp, South Korea (2023)
  • Recipient of PM's Laptop under the PM's Youth Laptop Scheme Phase III (2023)

Education

  • PhD in Electrical Engineering
    Stanford University
    Sep. 2025 - present
    Advisor: Prof. John Cioffi
  • B.E. in Electrical Engineering
    National University of Sciences & Technology (NUST)
    Oct. 2020 - June 2024
    CGPA. 3.97/4.00
    Rector's Gold Medalist
    Chancellor's Silver Medalist

Research Interests

  • optimization and control
  • AI/ML
  • intelligent communications
  • reinforcement learning
  • algorithmic reasoning

Technical Skills

  • Python
  • C/C++
  • MATLAB
  • Mathematica
  • PyTorch
  • JAX
  • RLlib
  • TorchRL
  • LangChain

Select Projects

b-gpm

Built a Bayesian extension of the general preference model for efficient preference elicitation in RLHF. Uses probabilistic embeddings with closed-form uncertainty estimates to support active learning through information-theoretic acquisition functions.

comyx

Created a modular library for wireless network simulation that includes a statistical framework based on gamma approximations for analytical modeling. Implemented Numba-based JIT optimizations and parallelization to accelerate simulations.

aecc

Developed a vision Transformer-based autoencoder for image compression, transmission, and denoising. Outperformed traditional methods with 99.3% SSIM and 38 dB PSNR.

rl-wireless

Built a multi-cell massive MIMO environment for dynamic resource allocation and network sum-rate optimization under practical constraints. Trained and evaluated various DRL agents including DQN, TD3, and R2D2.