AI/ML Engineer • Researcher • Freelancer
I'm an AI/ML Engineer and passionate problem-solver from Khyber Pakhtunkhwa, Pakistan. With a Master's degree in Artificial Intelligence and published research spanning computer vision, biomedical signal processing, and intelligent healthcare systems, I bridge academic research with real-world deployment.
My expertise spans machine learning, deep learning, computer vision, NLP, and MLOps . I've developed AI solutions across cybersecurity, smart cities, healthcare, disaster prediction, and IoT systems. Currently working as an AI/ML Engineer at Horizon Tech Services, where I develop intelligent systems that push innovation beyond traditional boundaries.
I enjoy working at the intersection of research and applied AI—transforming theoretical concepts into scalable products that solve real problems. I am particularly interested in advancing AI for human well-being, healthcare intelligence, and socially beneficial applications.
Actively seeking PhD opportunities in Computer Science, AI, ML, Computer Vision, Biomedical Engineering, NLP, or related fields to contribute to innovative and impactful research.
🎯 Core Expertise
Deep Learning & AI
CNN, RNN, LSTM, Transformers, GANs, Reinforcement Learning
Computer Vision
Object Detection, YOLO, Image Segmentation, 3D Detection, Vision Transformer
Natural Language Processing
LLMs, Text Classification, Chatbots, NLP Pipelines. OpenAI
MLOps & Deployment
FastAPI, AWS, Azure, Docker, Kubernetes, CI/CD
💼 Professional Experience
AI/ML Engineer – Horizon Tech Services
Aug 2025 – Present
LLM Inferencing and Optimization Tasks
Developed AI-driven cybersecurity systems with UEBA, designed 24+ anomaly correlation rules reducing false positives by 21%, and built scalable data pipelines using Apache Kafka and OpenSearch. Deployed microservices using Docker & Kubernetes with MLflow monitoring.
Sr. AI Engineer – CISNR, UET Peshawar
Aug 2024 – Jul 2025
Led development of AI-powered water contamination detection system, reducing detection time from 24 hours to 6 hours. Implemented evolutionary algorithms for self-learning chess AI, designed earthquake early-warning systems providing 5-second alerts, and developed flood prediction models achieving 85% accuracy.
Jr. AI Engineer – CISNR, UET Peshawar
Oct 2022 – Feb 2024
Developed NLP chatbots and conversational AI systems. Built FastAPI web applications deployed with Docker. Contributed to Safe City Project by integrating person & vehicle detection models into surveillance networks, improving public safety across Mardan's 20+ cameras.
🎓 Education
Master of Science in Artificial Intelligence
University of Engineering and Technology Peshawar (2023 – 2025)
CGPA: 3.84/4.0
Thesis: AIoT-Based Kit for Early Detection of Escherichia coli and Klebsiella Aerogenes Bacteria in Drinking Water
Bachelor of Science in Electrical Engineering
University of Engineering & Technology Peshawar (2018 – 2022)
CGPA: 3.27/4.0
Thesis: Designing a model to detect and classify Arrhythmias from ECG signals using Deep Learning
🔬 Research Interests
Machine Learning & Deep Learning
Biomedical Signal Processing & Healthcare AI
Computer Vision & Image Processing
Natural Language Processing & LLMs
Intelligent System & Autonomous Systems
IoT & Edge Computing for Smart Systems
Cybersecurity & Anomaly Detection
AI Ethics & Explainable AI
🏆 Featured Projects
Web-Based CRM & Learning Management Portal
Full-stack CRM + LMS system for teachers, students, and admins
Node.js REST APIs for user management and exam workflows
AI-based exam simulation, auto-feedback & question generation
Performance analytics and digital certificates/badges
Dockerized deployment for secure, scalable multi-user access
Tech: Node.js, REST APIs, Docker, AI Question Generation, Predictive Analytics, Digital Credentials
Safe City AI Surveillance
Gun detection, violence recognition, facial identification, and real-time alerting across 20+ cameras. Deployed in Mardan, Pakistan with sub-2s latency facial recognition integrated with police database.
Tech: YOLOv5, Deep Learning, Kubernetes
AI Water Contamination Detection
Reduced E. coli detection from 24 hours to <6 hours using microscopic image enhancement and real-time ML+CV systems. UNICEF-funded IoT-based early detection kit for drinking water safety.
Tech: OpenCV, TensorFlow, IoT, Real-time Processing
Early Flood Prediction & Alert System (Kalam & Swat Region)
Developed an AI-driven early flood forecasting system trained on Australian hydrological datasets and further fine-tuned for real-time conditions in Kalam and Swat, Pakistan. Utilized water-flow rate, rainfall forecasts, and climate indicators to predict flood onset, intensity, and spread with advance warning. Integrated real-time data pipelines to provide early alerts to communities, enabling timely evacuation and preventive safety measures.
Tech: AI Forecasting, Hydrological Data, Real-time Pipelines, Early Warning Systems
AI-Powered Earthquake Early Warning System (Jetson Nano + Seismometer Sensor)
Built a real-time earthquake early warning system deployed on NVIDIA Jetson Nano for on-device edge inference. Used a Seismometer to capture early seismic activity and differentiate P-waves from destructive S-waves. Implemented a hybrid ANN + LSTM model to analyze P-wave signals and trigger alerts ~10 seconds before S-waves arrive, providing critical life-saving response time. Designed for low-latency decision-making without cloud dependency, ensuring alerts remain operational during disasters.
Tech: NVIDIA Jetson Nano, Seismometer, ANN + LSTM, Edge Computing, P-wave Detection
Autonomous Vehicle 3D Detection
Trained and optimized 3D object detection models achieving mAP=90%. Converted Pandaset to KITTI format, compared PointRCNN, PV-RCNN, SECOND, CenterPoint for vehicle/pedestrian detection.
Tech: YOLO3D, PointPillars, Sensor Fusion
Self-Evolving Chess Bot
Autonomous agent capable of improving strategy through evolutionary algorithms without explicit rule-based programming. Demonstrated emergent AI behavior using genetic mutation, crossover, and tournament selection.
Tech: Evolutionary Algorithms, Genetic Programming
Face Emotion Recognition
Multi-modal emotion detection from image and speech using ResNet50 and MobileNetV2. Detects emotions: Happy, Sad, Surprise, Fear with real-time processing.
GitHub: View Repository
OCR with Bounding Boxes
Optical Character Recognition system that automatically creates line-level bounding boxes in scanned documents. Enhances usability by converting printed/handwritten text to machine-readable format with precise localization.
GitHub: View Repository
Face Recognition Attendance System
Automated attendance system leveraging real-time facial recognition for accurate and efficient employee or student tracking. Integrates with existing databases for seamless record keeping.
Tech: OpenCV, Deep Learning, Biometrics
Desktop Application for Bankruptcy Prediction
Predictive model integrated into a user-friendly desktop application to assess the likelihood of corporate bankruptcy based on financial indicators. Aids in risk management and decision-making.
Tech: Machine Learning, Python, GUI Development
Image Stitching (Panorama)
A computer vision project that stitches multiple overlapping images together to create a seamless panoramic view. Features robust keypoint detection and image warping algorithms.
Tech: OpenCV, SIFT/SURF, Image Processing
Lane Detection
Real-time lane detection system for autonomous vehicles, identifying and tracking lane lines using advanced computer vision techniques. Essential for navigation and safety in ADAS.
Tech: OpenCV, Hough Transform, Image Filtering
Principal Component Analysis (PCA)
Implementation of Principal Component Analysis for dimensionality reduction and data visualization. Demonstrated effectiveness on complex datasets, highlighting key variances.
Tech: Scikit-learn, NumPy, Data Analysis
📚 Research Publications
IJIST 2024 | Vol. 6, pp. 338–351
- Proposed a modular multi-digit recognition system by decomposing complex numbers into individual digit predictions.
- Utilized multiple single-digit CNN models combined with segmentation and data augmentation and digit-isolation pipeline for improved generalization.
- Demonstrated high recognition accuracy of 99.28 while maintaining scalability for variable-length digit sequences.
IJIST 2024 | Vol. 6, pp. 207–215
- Introduced a novel CNN-based stress prediction framework capable of estimating continuous stress scores (0–100) from short ECG segments.
- Optimized signal duration and sampling frequency, identifying 5-second ECG data at 200 Hz as the most efficient configuration.
- Achieved 95.04% accuracy in binary stress classification and strong performance in multiclass stress recognition.
IEEE ICETECC 2025 | mAP@50: 99%
- Developed a real-time, vision-based gas leakage detection system to enhance home and industrial safety without specialized gas sensors.
- Applied computer vision and deep learning models to visual patterns of gas meter associated under varying lighting conditions.
- Demonstrated accurate and fast leakage detection, enabling early hazard identification and reduced response time. Achieved 98% recall and 97% F1 score for real-time monitoring.
IEEE ICETECC 2025 | mAP@50: 0.96
- Proposed a multi-plane MRI-based deep learning framework with explicit non-brain tissue removal to improve robustness and diagnostic accuracy.
- Employed preprocessing for skull stripping followed by CNN-based feature learning across axial, sagittal, and coronal MRI planes.
- Achieved higher classification accuracy and model generalization by eliminating irrelevant non-brain regions from MRI inputs.
IEEE ICET-25 | mAP@50: 0.97
- Field surveys conducted in Swat and Peshawar provide real-world freshwater samples for E. coli contamination analysis.
- Microscopic cameras combined with computer vision models enable automated detection of E. coli, reducing reliance on manual laboratory inspection.
- The proposed framework reduces contamination detection time from 24 hours to approximately 8 hours, significantly improving rapid water safety assessment.
NeuroMarker | Elsevier | 2026
- A robust 3D-CNN for brain age prediction is trained on multi-site MRI data from OASIS, IXI, ABIDE, and ABIDE II using minimal preprocessing.
- Systematic robustness experiments show that rotation augmentation and non-brain tissue removal substantially improve accuracy, stability, and model optimization.
- The proposed framework achieves high predictive performance (r = 0.90, RMSE = 3.66 years) and provides uncertainty-aware brain-age biomarkers for neurodegenerative disease research.
IJIST 2024 | Vol. 8, pp. 31029–1048
- End-to-end child safety surveillance pipeline integrating detection, tracking, and activity recognition
- Custom datasets: 19.8K images + 1.96K video clips (47 child activity classes)
- Best results: YOLO26s (98.5% mAP) and VideoMAE (86.38% accuracy, 90.49% F1-score) for real-time monitoring
AIoT-Based Kit for Early Detection of Escherichia coli and Klebsiella aerogenes Coliforms in Drinking Water
Water Science and Engineerining | Elsevier (Q1) | MS Thesis | Under Review
- AI+ IoT Portable Kit For early Coliform detection and classification.
- Lab and Field testing with portable kit and detection system for Coliform Detection in Khyber Pakhtunkhwa Pakistan.
- Combines microscopic image analysis with Vision Transformer and Edge divice for remote testing data for rapid identification.
Robust AI-SCORE Framework: Independent and Adversarial Validation for Malware Detection
Scientific Reports | Springer Nature | Under Review
- Introduced an AI-SCORE–based malware detection framework that ensures robust feature consistency against obfuscation and adversarial attacks.
- Achieved near-perfect detection and classification accuracy (up to 99–100%) across MS-Office and PE files using ML and CNN models.
- Validated strong generalization on unseen and adversarial malware, confirming real-world reliability through independent testing
🛠️ Technical Skills & Tools
ML & AI Frameworks
- TensorFlow, Keras, PyTorch
- Scikit-learn, XGBoost
- OpenCV, YOLO (v5–v12)
- Generative AI, LLMs
Data & Deployment
- Python (Expert), FastAPI, Flask
- Docker, Kubernetes, Git
- AWS, Azure, GCP
- MLflow, CI/CD Pipelines
Data Engineering
- Apache Kafka, Apache NiFi
- SQL, NoSQL, ETL Pipelines
- Pandas, NumPy, SciPy
- Elasticsearch, OpenSearch
Specializations
- Biomedical Signal Processing
- Medical Imaging (MRI, X-ray)
- IoT & Edge AI
- Cybersecurity (UEBA)
- AIoT
🎓 Certifications & Training
🏅 Awards & Recognition
🥈 Best Summer Intern (2021)
National Centre of Artificial Intelligence – UET Peshawar for outstanding research and technical skills in AI Healthcare.
🏆 District Quiz Competition (2015)
2nd Place in District Higher Schools Tournament, Shangla for academic excellence and competitive knowledge.
💼 Freelancing Services
AI/ML Model Development
Custom machine learning and deep learning solutions tailored to your business needs.
Computer Vision Projects
Object detection, image segmentation, face recognition, and real-time video processing.
Python & NLP Solutions
Chatbots, text analysis, language models, and natural language processing systems.
Data Engineering & MLOps
ETL pipelines, model deployment, cloud integration, and production ML systems.
📞 Get In Touch
Contact Information
Email: 📧 [email protected]
Alternative: 📧 [email protected]
Phone: 📱 +92 343 917 6731
WhatsApp: 📱 +92 343 917 6731
Location: Islamabad, Pakistan
Connect
With Me
Jehad Ur Rahman