hi! i'm ahmad faisal.
an AI engineer & researcher that loves all things related to math, CS and engineering.
specializing in deep learning, wireless communication and real-time systems
about
i graduated in NUST in 2025 as an electrical engineer, with a passion for AI/ML. i explore intersections between engineering and AI, where artificial intelligence can create meaningful impact through research and practical applications.
with extensive research experience and freelancing projects, i focus on fundamentals while staying eager to learn. currently i'm in montreal, canada - researching deep learning for wireless networks on the side until october!
technical skills
Interests
featured projects
some projects that i think are pretty fun.
Real-Time Pakistani Sign Language (MLOps)
End-to-end PSL classifier with FastAPI web UI + WebSocket realtime inference, packaged in Docker with DVC pipelines and CI/CD to AWS EC2.
Chest Cancer Classification (MLflow + DVC)
Production-oriented pipeline for chest cancer classification with experiment tracking (MLflow), data/version control (DVC), and GitHub Actions CI/CD to AWS.
RAG Medical Chatbot (Gemini + LangChain)
End-to-end RAG pipeline using Gemini 1.5 Pro with LangChain, Pinecone DB, and Flask UI for medical query assistance.
Real-Time Lip Reading (LipNet)
Webcam lip-reading using 3D-CNN + BiLSTM architecture trained with CTC loss; includes real-time inference script and training notebook.
Semantic Segmentation — Indian Driving Dataset
Comparative study across DL and classic Computer Vision techniques: UNet, Segformer, ResNet vs. Otsu's Method, K-means on challenging Indian road scenarios.
Automated Pancreas — RL + Predictive Modeling
Insulin feedback-loop prototype combining DL predictive model (LSTM) with prescriptive RL agent to regulate insulin delivery.
publications & research
academic contributions and research publications
Published
Vision-Aided Beamforming: Revolutionizing Spectrum Efficiency for 6G Networks
Authors: Ahmad Faisal Mirza, Tayyib Ul Hassan, Syed Asad Ullah, Sajjad Hussain Chauhdary, Kapal Dev, Syed Ali Hassan
Published in: IEEE Network Magazine
Year: 2025
Explores vision-assisted beam prediction using real-world 6G datasets, highlighting single-frame beam index prediction for low-latency links.
Multi-Agent Deep Reinforcement Learning for Energy Efficient RIS-Assisted CoMP-NOMA Networks
Authors: Ahmad Faisal Mirza, Fatima Binte Tanveer, Muhammad Umer, Haejoon Jung, Syed Ali Hassan
Published in: IEEE International Conference on Communications (ICC 2025)
Year: 2025
A multi-agent deep reinforcement learning framework that jointly optimizes RIS-assisted CoMP-NOMA for improved energy efficiency.
Beyond Directional-RIS Aided NOMA-ISAC Networks: A DRL Approach for Sum-Rate Optimization
Authors: Ahmad Faisal Mirza, Ali Arshad Nasir, Haejoon Jung, Aamir Mahmood, Syed Ali Hassan, Mikael Gidlund
Published in: IEEE Communications Letters
Year: 2025
Formulates DRL-based joint optimization for BD-RIS aided NOMA-ISAC to maximize sum-rate under practical constraints.
Speaker Recognition: A Comparative Analysis Between Deep Learning and Non-Deep Learning Methodologies
Authors: Ahmad Faisal Mirza, Muhammad Mustafa, Zoha Ahmed, Sakhi Usman Akbar
Published in: The Asian Bulletin of Big Data Management
Year: 2024
Compares deep neural approaches with vector-quantization baselines for speaker recognition across accuracy and complexity.
A Deep Learning-Based Technique for Classification of Rice Leaf Disease Using Transfer Learning
Authors: Zainab Khalid, Huma Tauseef, Aqsa Shabbir, Naveed Iqbal, Sahar Zia, Ahmad Faisal
Published in: Kurdish Studies, Vol. 12 No. 2
Year: 2024
MobileNetV2-based transfer learning pipeline achieving high-accuracy rice leaf disease classification on a merged multi-source dataset.
A Machine Learning Approach for Textual Sentiment Classification
Authors: Huma Tauseef, Tehmina Shahid, Naveed Iqbal, Aqsa Shabbir, Sahar Zia, Ahmad Faisal Mirza, Sajjad Rabbani
Published in: The Asian Bulletin of Big Data Management, Vol. 3 No. 1
Year: 2023
Evaluates LSTM and BiLSTM (GloVe embeddings) for sentiment polarity on Amazon review datasets, with BiLSTM achieving the best accuracy.
Early Detection and Classification of Apple Leaf Diseases Using Deep Learning Technique
Authors: Areej Gul Khan, Rabia Javed, Ahsan Mukhtar Malik, Ahmad Faisal Mirza, Shazia Aslam, Aqsa Shabbir, Huma Tauseef
Published in: The Asian Bulletin of Big Data Management, Vol. 4 No. 3
Year: 2024
Uses VGG16 to identify early-stage apple leaf symptoms; shows improvements over standard baselines for timely intervention.
Accepted
AI-Driven Dynamic Spectrum Management in Integrated TN-NTN Networks
Authors: Ahmad Faisal Mirza, Muhammad Umer, Syed Ali Hassan
Published in: Book chapter in "Artificial Intelligence for Integrated Terrestrial and Non-Terrestrial Networks" (ed. Ali Jamshed)
Year: 2025
Surveys AI-enabled spectrum management strategies for integrated terrestrial and non-terrestrial networks, emphasizing efficiency and reliability.
DRL Based Phase Shift Optimization in Uplink SIM Under Hardware Constraints
Authors: Hafiz Badar Ur Zaman, Muhammad Abubaker, Manahil Ahmad, Ahmad Faisal Mirza, Syed Ali Hassan
Published in: IEEE GLOBECOM 2025 (accepted)
Year: 2025
A deep reinforcement learning approach for uplink SIM phase-shift optimization under practical hardware constraints.
In Progress
DRL for Multi-Stage Optimization of BD-RIS Assisted NOMA-ISAC Networks, A Novel Approach
Authors: Ahmad Faisal Mirza (first author), et al.
Published in: Manuscript in progress
Year: —
Work-in-progress on a multi-stage DRL framework for BD-RIS aided NOMA-ISAC, targeting joint sensing-communication trade-offs.
Modelling and Optimization of a Differentiable Digital Twin for Optimization of Discrete RIS Spaces under Hardware Constraints
Authors: Ahmad Faisal Mirza (first author), et al.
Published in: Manuscript in progress
Year: —
Developing a differentiable digital twin to navigate discrete RIS configuration spaces and hardware constraints via gradient-friendly relaxations.
Submitted
End-to-End Learning with EM-Aware Differentiable-Ready Discrete-Phase RIS for MIMO-OFDM via Ray Tracing
Authors: Ahmad Faisal Mirza, Messaoud Ahmed Ouameur, Mohammed Ali Dou, Miloud Bagaa, Muhammad Ahmad Mohsin, Sinem Coleri, Daniel Massicotte
Published in: IEEE Transactions on Green Communications and Networking (submitted)
Year: 2025
Proposes an EM-aware differentiable ray-tracing framework with 1-bit discrete RIS control and an end-to-end learned neural receiver for MIMO-OFDM links.
awards & grants
recognition for academic excellence and research contributions in AI and wireless communications
MITACS GRI 2025 Scholarship
2025Awarded MITACS GRI 2025 Scholarship for the project "Intelligent reconfigurable surfaces enabled communication networks" at Université du Québec à Trois–Rivières, Quebec, Canada under the supervision of Dr. Messaoud Ahmed Ouameur.
IIPP Taiwan 2025 Scholarship
2025Awarded IIPP Internship 2025 for the project "Advanced Reconfigurable Intelligent Surface (RIS)-Assisted MIMO Systems for robust secure communication" at National Chung Cheng University, Minxiong, Chiayi, Taiwan under the supervision of Dr. Jen-Yi Pan.
DAAD Scholarship
2024Awarded DAAD Scholarship for the project "Transformers for FPGA" at RPTU Kaiserslautern, Germany under the supervision of Dr. Christian Weiss.
Exchange Programme - Sapienza University of Rome
2024One of two bachelor students selected for the exchange programme at Sapienza University of Rome from a batch of 200.
Academic Scholarship
2021-2025Awarded scholarship for excellent academic performance in 1st and 5th semester, awarded to top 2.5% of students in a batch of 200.