AI, engineered for healthcare.
Most medical AI never leaves the paper it was published in. I build the systems that do, closing the gap between what a model can do in research and what it actually changes at the bedside, turning models into systems clinicians trust and patients benefit from.
The most capable model in the world means nothing if it never reaches a patient.
I sit at the intersection of engineering and medicine, turning powerful models into care that’s real, safe, and trusted.
Most medical AI stalls in papers and demos. I take on the harder work: turning it into systems clinicians actually rely on, not just cite.
Every design decision traces back to one person waiting on an answer. Safety and clarity always come first, no exceptions.
Regulatory constraints and clinical validation aren’t bolted on at the end. They’re design inputs from line one of code.
Three degrees in biomedical engineering taught me to carry an idea from concept to a system people can actually depend on.
Open-source work spanning the AI stack: agents, retrieval, 3D segmentation, classification, and the tooling to ship trained weights. Each is a standalone, tested repository, and each reports the metric that matters rather than the one that flatters.
A dependency-light framework for tool-using LLM agents: ReAct loop, task planner, tool registry and bounded memory. The LLM sits behind a two-method interface, so the whole control flow is tested offline with a scripted client, no API key. Swapping in Claude is a one-line change.
RAG over clinical documents, built so dosages and lab values survive retrieval intact, the failure mode that breaks naive RAG in medicine. Declines to answer when retrieval confidence is low, and runs fully local so no patient data leaves the machine.
A 3D Attention U-Net segmenting tumor core, whole tumor and enhancing tumor from 4-modality MRI. Split by patient, evaluated on full volumes with sliding-window inference, scored on both official BraTS metrics: mean Dice 0.8069.
EfficientNet-B0 on HAM10000, split by lesion to stop leakage. It reports 83.1% accuracy and 0.965 macro AUC, then shows what those hide: 64.4% sensitivity on malignant lesions. Browser demo runs on ONNX, so the image never leaves the device.
Model cards, a machine-readable manifest, and download and verify tooling for the trained checkpoints above, hosted on Hugging Face with GitHub release mirrors. Verification fingerprints the architecture, not just the file size.
Core competencies across medical imaging, simulation, and regulated AI deployment, each one earning its place in a clinical system.
MRI-to-ultrasound registration and biomechanical modeling, built to survive an actual OR, not just a benchmark.
Segmentation, registration, and full 3D pipelines: MRI, CT, ultrasound, whatever the modality demands.
Retrieval and multi-agent orchestration for clinical knowledge, where a wrong answer isn’t bad UX. It’s a risk.
Graph learning constrained by physical law, anatomy and tissue modeled as structure, not flat features.
Finite-element tissue deformation for surgical planning, and physics-accurate synthetic data when real data can’t scale.
Production pipelines, reinforcement learning for sequential decisions, and deployment that survives FDA scrutiny.
Clinical NLP for EHR text, and biosignal processing built for the noise profile of real medical devices.
FDA SaMD pathways, 510(k)/PMA awareness, and clinical validation for AI that ships as a device, not a demo.
Sensor and system calibration, accuracy testing, and uncertainty quantification. It’s the discipline behind that 4.6 mm number.
Python, PyTorch, TensorFlow, CNNs, GNNs, RNNs, Transformers & Attention, Random Forests, Gradient Boosting (XGBoost), SVM, Bayesian Methods, Reinforcement Learning, Generative AI, LLM & Agentic AI, RAG Engineering, LangChain, Vector Databases, Prompt Engineering, Fine-Tuning & Quantization, NLP, HPC, CUDA
DevOps / DataOps / MLOps, Azure ML, Azure Databricks, AWS SageMaker, GCP Vertex AI, Docker, Kubernetes, Terraform, CI/CD, MLflow, Airflow, Model Monitoring, Feature Stores, Data Engineering, Microservices
Segmentation, Registration, DICOM, NIfTI, MINC, MRI 3D, Ultrasound, 3D Slicer, MONAI, ITK/VTK, PLUS Toolkit, Radiomics, Multi-Modal Fusion
Abaqus, ANSYS, FEA, Digital Twins, CGAL, Blender, MeshLab, Patient-Specific Meshing, 3D Printing
Surgical Planning, Surgical Tracking, Optical & Electromagnetic Tracking, Intraoperative Imaging, Registration Accuracy Assessment, Image-Guided Therapy, Soft-Tissue Modeling
Git & Version Control, Agile / Scrum, JIRA, MS Visual Studio, CMake, DOE, ANOVA, Minitab, LabVIEW
FDA Submissions, SaMD, 510(k), Biocompatibility, EU MDR/IVDR, HIPAA, GDPR, Clinical Validation, AI in MedTech
MS Word / Excel / PowerPoint, Visio, EndNote, Teams, SharePoint, Slack, Google Workspace
Ongoing coursework to stay current across the fast-moving parts of the AI stack, from agentic systems to cloud MLOps.
Four-course sequence: Fundamentals of RL, Sample-based Learning Methods, Prediction & Control with Function Approximation, and a capstone building a complete RL system.
Built CNN models for image classification and segmentation to diagnose lung and brain disorders from medical scans.
Introduction to TensorFlow, CNNs in TensorFlow, and NLP in TensorFlow, building scalable models for image and text data.
Foundations of building applications on top of generative AI models.
Designing and deploying autonomous AI agents on Microsoft’s Azure AI Foundry platform.
End-to-end model pipelines and data science workflows on Azure’s cloud ML stack.
Production practices for shipping and maintaining ML systems reliably at scale.
Data gathering, missing-data handling, feature extraction and selection using PCA and variance thresholds.
Deeper architectures and training techniques beyond introductory ML.
Building virtual representations of physical systems for simulation-driven design and monitoring.
Applied simulation workflows using the ANSYS platform.
Core ML theory and applied methods.
Academic conferences from my Ph.D. research, each resulting in a peer-reviewed publication, alongside recent industry events tracking where AI, simulation, and medical-device regulation are heading.
I’m a biomedical engineer with a Ph.D., now Senior AI/ML Engineer at Abbott, building AI and machine learning for medical imaging and healthcare at a company that ships to actual patients.
My doctoral research built a real-time, image-guided navigation system for breast cancer surgery: deep learning, patient-specific modeling, and biomechanical simulation combined in one framework, reaching 4.6 mm tumor localization accuracy, developed alongside clinical teams from diagnosis through treatment planning.
Three degrees across Cairo, Connecticut, and Virginia, and hands-on work in medical imaging, biomechanics, bioelectric engineering, and drug delivery. That breadth is exactly what lets me move research into clinical practice instead of leaving it in a paper.
Today I lead AI integration for my department at Abbott: setting technical direction, establishing the team’s AI/ML strategy, tooling and engineering standards, and mentoring the engineers who build alongside me. I’m also completing an MBA at the University of Texas at Dallas, to pair the engineering with business strategy.
Views on this site are my own, not my employer’s.
Peer-reviewed conference work in medical imaging, breast MRI segmentation, and finite-element surgical simulation.
M. Alqaoud, J. Plemmons, E. Feliberti, S. Dong, K. Kaipa, G. Fichtinger, Y. Xiao, M. A. Audette
TLDR: A deep-learning pipeline segmenting multi-modality breast MRI with nnU-Net, paired with a tissue-delineating phantom, to support planning for robotic tumor surgery.
View on IEEE Xplore ↗M. Alqaoud, J. Plemmons, E. Feliberti, K. Kaipa, S. Dong, G. Fichtinger, Y. Xiao, M. A. Audette
TLDR: Extends multi-modality breast MRI segmentation to preoperative planning for robotic surgical navigation. Best Paper, Medical Track.
View on IEEE Xplore ↗M. Alqaoud, J. Plemmons, E. Feliberti, K. Kaipa, G. Fichtinger, Y. Xiao, T. Rashid, M. A. Audette
TLDR: A controlled-resolution, multi-material breast meshing method enabling finite-element-based interactive surgical simulation guided by the surgical approach.
View on IEEE Xplore ↗M. Alqaoud, M. A. Audette, et al.
TLDR: A biomechanical simulation modeling breast tissue deformation caused by ultrasound probe pressure, improving the accuracy of MRI-to-ultrasound registration for surgical navigation. Presented at ANNSIM 2025, Madrid, Spain.
View on IEEE Xplore ↗Motaz Alqaoud · Old Dominion University
TLDR: The full doctoral dissertation: an end-to-end AI-driven navigation system integrating deep learning, patient-specific modeling, and biomechanical simulation for breast cancer surgery, achieving 4.6 mm tumor localization.
View on ODU Digital Commons ↗I’m always glad to hear from people working on AI in healthcare, potential collaborators, or anyone with a sharp question about the work above.