About Me
Location: Atlanta, GA
Email: rafiei.ac@gmail.com
LinkedIn: mhrafiei | Portfolio: ep.jhu.edu/faculty/mohammad-rafiei
Inventor of the Neural Dynamic Classifier, Structural Health Index model, accelerated DD approximation, and MOFHEI. Passionate about bridging academic research with real-world AI solutions.
Professional Summary
● 15+ Years of ML/AI model development in Python (TensorFlow, PyTorch).
● R&D Leadership: 8+ years leading projects in self-supervised learning, privacy-preserving ML, HPC-based optimization.
● Publications & Grants: 23+ peer-reviewed ML/AI articles (3,100+ citations), experience with NAS, NSF, DOD, DARPA proposals.
● Academic Teaching: Part-time instructor at Johns Hopkins University, developed online courses (self-supervised.com & python3h.com).
● Startup Experience: Founded AIWhittler focusing on model parameter reduction for deep learning efficiency.
● Areas of Focus: NLP & LLMs, DevOps/MLOps, HPC & distributed GPU, signal processing, BCI (EEG/ECG), computer vision, adversarial & secure AI.
Technical Skills
Intermediate to Proficient
Core Tools & Libraries
- Python 3+, Shell Script, MATLAB
- TensorFlow, Keras, PyTorch, HuggingFace
- Scikit-Learn, NumPy, Pandas, Matplotlib
- Docker, Kubernetes, GHCR
- CI/CD, Git/GitHub, VSCode, Anaconda
- Colab, Kaggle, Lambda Labs, Paperspace
- Gcloud, Boto3, AWS (S3, EC2), GCP (VMs, TPU)
- Distributed GPU, HPC, Supercomputing
- OpenAI API, Gemini API
- OpenCV, Pillow, MoviePy, ffmpeg
Domains & Concepts
- ML, AI, Constraint Optimization, Grants & Proposals
- Self-Supervised & Representation Learning, Contrastive
- GANs: StyleGAN, Pro-GAN, Diffusion
- NLP, Attention, Transformers, LLMs, Prompt Eng.
- Computer Vision: R-CNN, SSD, YOLO, OCR
- Signal Processing: EEG, ECG, fMRI
- Knowledge Distillation, ML Compression, Pruning
- Adversarial Training & Defense, Explainable AI
- Privacy-Preserving AI, Homomorphic Encryption (HE)
- Responsible/Trustworthy AI, Ethics, MLOps, DevOps
Misc. Expertise
- Anomaly Detection, Unbalanced Data, Augmentation
- Random Forest, SVM, KNN, Naïve Bayes, GB, LR, Ensemble
- ResNet, Inception, EfficientNet, VGG, MobileNet, ConvNet
- Mechanical, Materials & Biomedical Eng., Neuroscience
- Tabular, Unstructured, Multi-Modal, Image, Text, Video
- Brain-Computer Interface, Time & Freq. Feature Extraction
- Detectron, Torchvision, Torchtext
- OOP, BeautifulSoup, MQL, Flask
- VSCode, HPC Shell, DockerHub
Beginner
Familiar Tech & Concepts
- SQL, LangChain, Azure, Reinforcement Learning
- HTML, CSS, JavaScript
- Scala, Spark
- Lambda, SageMaker
- Speech & Sound Recognition
- Django
- Cloudflare, Snowflake
- Graph Neural Networks
- DevSecOps
Featured Projects
Neural Dynamic Classification
A novel ML algorithm leveraging dynamic neural states for robust classification. Successfully applied in structural health monitoring, medical data analytics, and beyond, achieving high accuracy even in noisy environments.
Structural Health Index
An unsupervised deep learning framework for both global and local health assessment of complex structures. Introduces a new SHI metric derived from advanced frequency-domain analysis—reduces reliance on costly physical tests.
AI Whittler
A startup focusing on ML parameter reduction through the “Supervised Whittler” service. Enables efficient deep learning model pruning and refinement, tested on distributed GPUs for scalability and user-friendly deployment.
MOFHEI
A privacy-preserving MLaaS solution using TensorFlow and Homomorphic Encryption. Ensures data and model owners can collaborate for secure inference on untrusted systems while maintaining full confidentiality.
SimCLR for EEG/ECG
A self-supervised learning pipeline harnessing consumer EEG and ECG data for near-wild cognitive estimation. Employs domain-specific augmentations to boost accuracy when labeled data is scarce.
Self-Supervised Learning Course
Udemy-based training at self-supervised.com covering modern SSL frameworks (SimCLR, contrastive approaches) and domain adaptation strategies in real-world ML.
Python3h Course
Online Python 3 foundations offered at python3h.com. Blends engineering, data science, and advanced AI fundamentals, delivering a practical learning path for students.
Neuro Rehab ML Pipelines
Under rxgames.com, implemented advanced AI pipelines that process multi-sensor motion-capture data for stroke/MS patient therapy. Delivers personalized rehab solutions using data-driven analytics.
Accelerated DD Approximation
A cutting-edge technique for Discrete Dislocation (DD) simulations, drastically reducing computational overhead. Predicts stress fields and interaction forces more efficiently, revolutionizing materials engineering workflows.
Hobbies & Interests
I’m a passionate hiker and nature enthusiast. Spending time outdoors provides fresh perspectives, fuels my creativity, and balances the rigors of ML/AI research.
Exploring mountain trails
Taking in scenic vistas
Finding calm near rivers and lakes
Enjoying quiet moments in nature
Contact
Email: rafiei.ac@gmail.com
LinkedIn: @mhrafiei
GitHub: @mhrafiei