I'm a Senior AI Engineer with 5+ years of experience delivering end-to-end machine learning solutions across medical imaging, computer vision, natural language processing, and cloud-based AI platforms.
Currently, I am driving innovation at PaxeraHealth (Boston, MA/Remote), specializing in Medical Generative AI, 3D Segmentation, and Vision-Language Models.
- 🎓 BSc. Electronics & Communications Engineering (Cairo University, 2021)
- 📍 Location: Cairo, Egypt
- 🏥 Specialization: Medical AI, 3D Tumor Segmentation, RAG & VLMs
- 🔬 Research Focus: WiFi Sensing (CSI), Deep Learning, Edge AI
| Company | Role | Duration | Focus |
|---|---|---|---|
| PaxeraHealth | Senior AI Engineer | Nov 2023 - Present | Medical GenAI, 3D Segmentation (Kidney/Pancreas), MLOps |
| DevisionX | Senior CV Engineer | Apr 2022 - Jan 2024 | AutoML Pipelines, SAM Integration, Vision Transformers |
| PassApp | Computer Vision Engineer | Jun 2021 - Jan 2022 | Face Matching, LPR, Mobile Edge Deployment |
| Pxivision | Co-founder / ML Engineer | Dec 2020 - Apr 2022 | Entrepreneurship, Custom OCR & Surveillance Solutions |
| DevisionX | CV Engineer (Intern) | Jun 2020 - Aug 2020 | Object Detection (YOLO), Early AutoML Research |
- LLMs & VLMs: Llama 3, Mixtral, MedGemma, PaliGemma, Florence-2
- Serving: VLLM, SGLang, Ollama
- Fine-tuning: LLaMA-Factory, Unsloth, PEFT/LoRA
- Vector DBs: FAISS, Chroma, Pinecone
- Agents: CrewAI, LangChain, LlamaIndex
- 3D Processing: VTK, ITK, Monai, Open3D
- Medical Standards: DICOM, NIfTI
- Segmentation: SAM, SAM2, nnU-Net, Custom 3D CNNs
🏆 Blaze-Wi Research Project - WiFi Sensing & Deep Learning
- Objective: Replace invasive cameras with WiFi signals (CSI) for privacy-preserving surveillance.
- Outcome: Achieved 99.3% accuracy in gait recognition on the Widar 3.0 dataset.
- Impact: Developed real-time intrusion detection and indoor localization systems using Flutter and Deep Learning.
I write about advanced AI implementation on Medium.
- 🏥 Fine-Tuning MedGemma 4B on Free Colab
- Guide to fine-tuning Google's medical VLM for specialized diagnostic tasks.
- 🎯 Florence-2 VLM Fine-tuning on Custom Dataset
- Step-by-step tutorial for Microsoft's Florence-2 VLM.
- 🔬 Finetune PaliGemma Vision Language Model
- Complete guide to Google's PaliGemma for custom vision-language tasks.
- 🚪 Training YOLOv8: Door Detection for the Visually Impaired
- Building accessible AI solutions for social impact.
Note: The majority of my work involves proprietary software for enterprise and medical clients (PaxeraHealth, DevisionX). My public repositories showcase research experiments, open-source contributions, and educational tutorials.
Key Private Projects:
- Medical 3D Segmentation: Production APIs for Kidney/Tumor/Pancreas segmentation (1.2s latency).
- EraBot: RAG + VLM Chatbot for assisting radiologists.
- Tuba AutoML: Enterprise platform for automated computer vision model training.
