Hello, I'm
Mohamad Mahayni
I am a Biomedical Engineer specializing in machine learning, signal processing, and edge AI. I design and build end to end systems that turn raw sensor data into optimized, real world intelligent solutions for industrial and biomedical sensing applications.
About Me
I'm a Biomedical Engineer from Syria, now based in the Black Forest region of Germany, finishing my M.Sc. at Anhalt University of Applied Sciences with a GPA of 1.6.
My thesis at Hahn-Schickard is where my passions converge: I built a complete ML pipeline that processes plant ultrasonic emissions (18–120 kHz), converts them into CWT spectrograms, and trains a CNN autoencoder for anomaly detection — compressed to just 160 KB for edge deployment on MEMS sensors.
I came to Germany because it's a global pioneer in AI and medical technology. I want to work at the frontier where intelligent systems meet real-world healthcare and industrial problems.
Experience
Master's Thesis Researcher
Hahn-Schickard · Villingen-Schwenningen, Germany
ULTRA-SAVE Project — Plant Stress Detection via ML
- Developed a full-scale ML pipeline from acoustic signal processing and CWT image generation to training a CNN-based autoencoder for anomaly detection
- Engineered the model for edge devices using a custom Optuna multi-objective optimization, achieving a final model size of just 160 KB
- Achieved >95% across AUC, F1-score, accuracy, and a problem-specific anomaly metric on held-out test data drawn from over 43,000 events
- Validated cost-effective MEMS sensors against high-end industrial standards
Office Assistant
Jawad Veterinary Medicine · Damascus, Syria
- Analyzed pharmaceutical operations data to support business decision-making
- Produced clear visualizations and reports for non-technical stakeholders
- Contributed to early digitalization initiatives, improving data traceability and cross-team workflows
Projects
Unsupervised Anomaly Detection for Bio-Signals
Hahn-Schickard · ULTRA-SAVE Project
End-to-end ML pipeline converting 1D plant ultrasonic emissions (18–120 kHz) into 2D CWT spectrograms, then applying a CNN Autoencoder with custom Optuna optimization for edge deployment.
🔬 Live Pipeline Sandbox
Simulate raw plant acoustic signals (18-120 kHz) and see the edge anomaly detector classify stress states in real time.
Radiation Dose Estimation from Skin Images
Anhalt UAS · Course Project
AI-driven method to estimate radiation dose in breast cancer patients from dermoscopic RGB skin images. Compared ResNet-18, Random Forest regression, and classification approaches with Leave-One-Patient-Out cross-validation.
Skills
🧠 Machine Learning & AI
📡 Signal & Image Processing
💻 Programming & Tools
🌍 Languages
Education
M.Sc. Biomedical Engineering
Anhalt University of Applied Sciences
Köthen (Anhalt), Germany
GPA: 1.6 (German scale)
B.Sc. Biomedical Engineering
Damascus University
Damascus, Syria
The Biomedical Engineering Path
How a typical academic BME pathway maps to Mohamad's high-impact specialization in ML & edge-deployed biosensors.
Electronics & Instrumentation
Standard BME MilestoneMastering raw sensor signal conditioning, operational amplifiers, and microcontroller architectures to interface biological systems with digital circuits.
🎯 Mohamad's Alignment:
Years of intense rigor at Damascus University, deep-diving into microprocessors, medical electronics, and radiation physics.
Signal & Image Processing
Standard BME MilestoneApplying digital filtering, Fourier analyses, and wavelet transforms to clean noisy signals (ECG, EMG, EEG) and reconstruct clinical medical images (CT, MRI).
🎯 Mohamad's Alignment:
Developed Continuous Wavelet Transform (CWT) time-frequency spectrogram matrices and feature extraction algorithms for skin lesion dose estimations.
Clinical AI & Modeling
Standard BME MilestoneUtilizing statistical modeling, regression analysis, and machine learning to build diagnostic classifiers and predictive prognosis engines.
🎯 Mohamad's Alignment:
Graduated with a 1.6 GPA (German Scale) M.Sc. at Anhalt UAS, specializing in Deep learning courses, autoencoders, and Optuna HPO algorithms.
Edge Deployment & Embedded AI
Mohamad's Frontier PeakCompacting complex algorithms to run locally on low-cost, low-power microcontrollers directly at the biosensor site without cloud dependencies.
🚀 Master's Breakthrough:
Compressed a PyTorch CNN Autoencoder into 160 KB for edge sensing (95%+ AUC on plant acoustic ultrasonic emissions at Hahn-Schickard).
Recommendation Letters
"Mohamad demonstrated excellent analytical capabilities, successfully optimizing a complex CNN Autoencoder model down to 160 KB for resource-constrained edge MEMS sensors."
"A highly dedicated and exceptionally skilled student in signal processing and deep learning methodologies. His academic accomplishments are highly commendable."
"Throughout his academic curriculum, Mohamad consistently chose academic excellence and deep engineering focus, excelling in automatic control systems and clinical modeling."
Hobbies & Creative Outlets
I play recreationally — it's my favourite way to switch off and reset after deep work. (Try the mini-game: drag the ball into the hoop.)
I keep up with new AI research and tooling for fun — these are the channels I learn from most.
IBM Technology
Deep dives into hybrid cloud, AI, and enterprise tech concepts.
Visual Studio Code
Official updates, tips, and tutorials for the VS Code editor.
Nate Herk | AI Automation
Practical tutorials on AI agents, automation, and workflow design.
Google Cloud Tech
Cloud architecture guides, Kubernetes tips, and Vertex AI tutorials.
NetworkChuck
High-energy networking, IT, cybersecurity, and coding tutorials.
GitHub
Official home for GitHub developer tutorials, features, and events.
I unwind with science, tech, and AI podcasts like Huberman Lab, Lex Fridman, and Der KI-Podcast.
Huberman Lab
Paused
Get in Touch
I'm actively looking for opportunities in ML/AI applied to biomedical or industrial sensing. Let's build something that matters.
