Mohammad Hussain

M.Sc. Researcher in Electrical & Electronics Engineering
Koç University · Wireless Networks Laboratory
Advised by Prof. Sinem Coleri

Seeking PhD positions for Fall 2027 in Deep Learning for Wireless / 6G PHY
Email Scholar ORCID arXiv Profile GitHub LinkedIn CV
Mohammad Hussain Photo

About Me

I am a Master's researcher in Electrical and Electronics Engineering at Koç University (Istanbul, Türkiye), working in the Wireless Networks Laboratory under the supervision of Prof. Sinem Coleri.

My research lies at the intersection of Uncertainty-Aware Machine Learning, Safe Reinforcement Learning, and Next-Generation Wireless Communications (5G-Advanced / 6G & O-RAN). I am particularly interested in developing theoretical and algorithmic frameworks—combining deep reinforcement learning, online conformal prediction, risk-sensitive optimization (CVaR), and deterministic safety shields—that provide provable reliability and extreme energy/computational efficiency for ultra-low-latency and broadband slicing. My current work is done in collaboration with Turkcell and Opticoms on AI-driven O-RAN resource allocation.

Between my B.Sc. and M.Sc., I worked as a Medical Imaging and AI intern at Radiologics, building CT segmentation and detection pipelines for pediatric and skeletal imaging.

Previously, I earned my B.Sc. in Electrical and Electronics Engineering from Bilkent University (Ankara, Turkey) in 2025 on an 80% Merit Scholarship (Top 15% class ranking).

Research Interests

Safe & Uncertainty-Aware Reinforcement Learning Constrained DRL, exact-cardinality sampling, execution-aware safety shielding, and multi-timescale control
Conformal Prediction & Statistical Reliability Causal one-sided conformal bounds, Conditional Value-at-Risk (CVaR), worst-user tail latency guarantees for URLLC
Hierarchical O-RAN Control & Slicing Multi-timescale Non-RT / Near-RT RIC xApps & O-DU scheduling for dynamic eMBB–URLLC coexistence
Resource-Efficient DL & PHY/MAC Layer Lightweight neural architectures, CSI prediction & compression (GRU-Attn-DSLH), Massive MIMO, 3GPP TR 38.901

News

  • May 2026 📄 Our paper "Resource-Efficient CSI Prediction: A Gated Fusion and Factorized Projection Approach" was published in IEEE Communications Letters! Check out the IEEE Xplore article and open-source code.
  • Oct 2025 🎓 Started M.Sc. in Electrical and Electronics Engineering at Koç University (Wireless Networks Laboratory) under the supervision of Prof. Sinem Coleri, working on AI-driven O-RAN resource allocation with Turkcell and Opticoms.
  • Jul 2025 🏥 Joined Radiologics as a Medical Imaging and AI intern, building CT segmentation and detection pipelines.
  • Jun 2025 🎉 Graduated with a B.Sc. in Electrical and Electronics Engineering from Bilkent University (Top 15% rank, 80% Merit Scholarship).

Publications

Resource-Efficient CSI Prediction: A Gated Fusion and Factorized Projection Approach

Mohammad Hussain, Maedeh Adibag, Dilara Gurer, Gokhan Kalem, Kerim Serin, Sinem Coleri
IEEE Communications Letters, vol. 30, pp. 1979–1983, 2026.

Accurate Channel State Information (CSI) prediction is vital for dynamic MIMO systems but computationally demanding. We propose a lightweight predictor combining a GRU encoder with Luong attention, a bottleneck gated fusion module, and a Dimension-wise Separable Linear Head (DSLH). Evaluated on 3GPP TR 38.901 channels, the model achieves −13.84 dB NMSE with 26% fewer parameters and ~2.3× higher inference throughput than LinFormer baselines.

IEEE Xplore arXiv 2605.06578 Code Repository

Research & Engineering Projects

Featured · Research

Resource-Efficient CSI Prediction Benchmark

PyTorch implementation of the GRU-Attn-DSLH framework and multiple baseline architectures (LinFormer, Vanilla LSTM, Vanilla GRU) for 2×2 MIMO channel state prediction with complete QuaDRiGa 3GPP TR 38.901 dataset generation scripts.

PyTorch MATLAB QuaDRiGa 3GPP TR 38.901 MIMO
Bilkent FYP · Industry: KAREL

Edge Vision & Real-Time Sensor Monitoring Platform

Built a multi-channel analytics system deployable on NVIDIA Jetson or standard hardware, integrating five task-specific vision models (PyTorch, YOLO, OpenCV) behind a PyQt5 operator interface. Designed a UART ingestion pipeline for CO₂, temperature, humidity, pressure, and acceleration streams with configurable thresholds and consecutive-anomaly detection, plus a Flask and Socket.IO data layer streaming live metrics between edge devices, a web dashboard, and a desktop client.

NVIDIA Jetson YOLO OpenCV PyQt5 Flask / Socket.IO
Robotics & ML

Neural Networks for Legged Locomotion

Individual research with Prof. Omer Morgül on modeling SLIP/TDSLIP dynamical systems. Analyzed variable interactions via custom phase-space visualizations and benchmarked GPR, SVM, and deep neural nets for apex prediction.

MATLAB Deep Learning Gaussian Processes Nonlinear Dynamics
Signal Processing

Uniform & Non-Uniform Scalar Quantization

Implemented Lloyd-Max companders and non-uniform scalar quantization routines in MATLAB for grayscale image compression and signal fidelity optimization.

DSP Lloyd-Max MATLAB Image Compression

Education

Koç University

Oct 2025 – Present

Master of Science in Electrical and Electronics Engineering

GPA: 3.57 / 4.00 · Wireless Networks Laboratory · Advisor: Prof. Sinem Coleri
Focus: Safe & uncertainty-aware RL, conformal prediction, resource-efficient deep learning for wireless communications and O-RAN.

Bilkent University

Graduated Jun 2025

Bachelor of Science in Electrical and Electronics Engineering

GPA: 3.27 / 4.00 · Class Ranking: Top 15%
Honors: 80% Merit Scholarship throughout entire undergraduate studies.
Core coursework: Digital Signal Processing, Telecommunications, Linear System Theory, Machine Learning, Embedded Systems.

Experience

Graduate Research Assistant · Koç University

Oct 2025 – Present

Wireless Networks Laboratory · Advisor: Prof. Sinem Coleri

  • Developing AI-driven O-RAN resource-allocation systems with Turkcell and Opticoms, using 3GPP TR 38.901 and QuaDRiGa channel data for model training and evaluation across LOS, NLOS, and mixed-condition scenarios.
  • Designed a GRU encoder with attention, gated feature fusion, and a factorized projection head, reducing parameters by 26% and increasing inference throughput by 2.3× over LinFormer while achieving −13.84 dB NMSE.
  • Building a hierarchical scheduling pipeline combining learned surrogate models, conformal demand bounds, convex optimization, and calibrated execution-time safety checks for latency-sensitive URLLC/eMBB workloads.

Medical Imaging and AI Intern · Radiologics

Jul 2025 – Oct 2025
  • Built CT processing and nnU-Net v2 segmentation pipelines for 44 anatomical structures, from DICOM conversion through geometry-preserving export.
  • Developed skeletal-muscle segmentation workflows for 90 L3-centered CT volumes with tissue- and bone-aware mask reconstruction.
  • Trained YOLOX detectors on 50K pediatric radiographs, improving recall by 5–10 points with domain-robust augmentations.
  • Built a QCT bone-density pipeline for artifact removal and HU-to-BMD calibration using Deming regression and Bland–Altman analysis.
  • Automated rotated bounding boxes for 21 anatomical regions, optimizing 30+ geometric parameters over 300+ Optuna trials per rule.

ML / Embedded Systems Intern · Modern Technology Laboratories

Aug 2024 – Sep 2024
  • Deployed a lightweight CNN inference pipeline on an ESP32 for resource-constrained structural crack detection.
  • Built PyTorch regression workflows for concrete-strength prediction, including preprocessing, training, and holdout evaluation.

Trainee Engineer · Nanomagnetic Instruments

Jun 2023 – Sep 2023
  • Configured AK4495 DACs and verified multi-channel ADC/DAC printed circuit boards.
  • Implemented I²S communication between Raspberry Pi Pico and FPGA boards for low-noise audio waveform synthesis.

Technical Toolkit

Machine Learning & AI

PyTorch, TensorFlow, Scikit-Learn, NumPy, SciPy, Matplotlib, Pandas, Hugging Face, RAG systems, Reinforcement Learning, Transformers, GRU/LSTM, Weights & Biases

Wireless & Simulation

QuaDRiGa Channel Modeling, 3GPP TR 38.901, MATLAB Wireless Toolbox, MIMO Simulation, 5G EVB

Edge, Vision & Medical Imaging

NVIDIA Jetson, ESP32, TensorFlow Lite, OpenCV, DICOM Processing, SimpleITK, YOLOX, Flask, Socket.IO

Programming & Hardware

Python, C/C++, SQL, MATLAB, VHDL, Bash, LaTeX, Raspberry Pi, FPGA (Basys 3), Git, Linux