Machine Learning Engineer | Computer Vision | Applied AI Systems
Riyadh, Saudi Arabia
I build applied AI systems around complex real-world problems, with experience spanning medical imaging, computer vision, dataset engineering, model evaluation, workflow design, and automation.
My work typically starts by structuring the problem and its data flow, then moves through implementation, controlled experimentation, evaluation, and iterative refinement.
Deep learning research for pelvic MRI-to-synthetic-CT generation in radiotherapy planning.
- Curated and quality-controlled paired MRI/CT cohorts with reproducible, frozen data splits
- Fine-tuned Med2Transformer, improving male validation MAE from 73.1 → 65.1 HU (~11%)
- Improved female zero-shot MAE from 261.5 → 102.0 HU, followed by female-specific fine-tuning to 87.4 HU
Focus: Medical Imaging · PyTorch · Synthetic CT · Fine-Tuning · Generalization · Error Analysis
Computer vision research for automated anatomical landmark localization in lateral cephalometric X-rays.
- Worked with 600 annotated X-ray images and 15 anatomical landmarks
- Trained and evaluated YOLO11 and HRNet-W48
- Built a preprocessing, training, inference, and landmark-level error-analysis pipeline with physical-distance interpretation
Focus: Computer Vision · PyTorch · YOLO11 · HRNet-W48 · X-ray Imaging · Landmark Localization
Designed a multi-role platform by translating a complex operational process into explicit roles, states, permissions, transitions, and traceable handoffs.
- Modeled the process as a stateful workflow rather than disconnected forms
- Designed role-based interactions across applicants, reviewers, approvers, auditors, and administrators
- Implemented request tracking, session management, access control, approval logic, and workflow traceability
Focus: System Design · Workflow Modeling · RBAC · PostgreSQL · Node.js · REST APIs
Built an end-to-end monitoring workflow connecting physical sensing, cloud data, validation logic, logging, and automated actions.
- Integrated ESP32-based sensing with Firebase
- Built n8n workflows for validation, threshold decisions, Google Sheets logging, and automated email alerts
- Separated sensing, storage, decision logic, automation, and notification into clear system components
Focus: ESP32 · Firebase · n8n · Automation · Cloud Integration · Event-Driven Workflows
Medical Imaging Research Platform — Designed and developed an interactive platform supporting medical-image discovery, MRI/CT review, quality control, human curation, experiment preparation, and model-result analysis.
Implementation details remain private while intellectual-property registration is in progress.
I tend to work from the structure of the problem outward:
Understand the workflow → identify constraints and decision points → structure the data → define quality gates → build the system or experiment → evaluate failures → refine
I am particularly interested in problems where AI, software systems, data, and real operational workflows intersect.
Machine Learning & AI
Python · PyTorch · CNNs · Transformers · GANs · Fine-Tuning · Transfer Learning · Model Evaluation
Computer Vision & Medical Imaging
OpenCV · MRI/CT · X-ray Imaging · Synthetic CT · Image Registration · Landmark Localization
Systems & Data
Workflow Design · Process Decomposition · PostgreSQL · Node.js · REST APIs · Firebase · Data Quality Control
Automation & Tools
n8n · Jupyter Notebook · Google Colab · Git
- LinkedIn: linkedin.com/in/yahyaalfar
- Email: yahya.a.alfar@gmail.com