I'm a Senior AI Engineer with 6+ years of experience building production AI systems across medical imaging, computer vision, generative AI, and multimodal applications.

I focus on the full lifecycle — from problem definition and architecture, through model selection and on-premise deployment — turning ambiguous problems into reliable, cost-efficient AI platforms that reach real users.

• 🧭 Currently: Senior AI Engineer at the Abu Dhabi Executive Office (ADEO), delivering strategic AI programs end-to-end
• 🏥 Previously: Led the AI team at PaxeraHealth, owning the technical direction of medical GenAI and computer-vision products
• 🎓 BSc. Electronics & Communications Engineering (2021)
• 📍 Location: Abu Dhabi, UAE
• 🔧 Focus: AI Platforms & Infrastructure, Agentic Systems, VLMs, 3D Medical Segmentation, RAG & GraphRAG, Voice AI
• 🔬 Research: WiFi Sensing (CSI), Deep Learning, Edge AI

• LLMs & VLMs: Llama 3, Qwen, Mixtral, Phi, MedGemma, PaliGemma, Florence-2
• Serving & Inference: vLLM, SGLang, llama.cpp, Ollama
• Fine-tuning: LLaMA-Factory, Unsloth, PEFT/LoRA
• Audio / ASR: Whisper, Parakeet, speech-to-speech pipelines
• Vector & Graph DBs: FAISS, Chroma, Pinecone; Neo4j (knowledge graphs, GraphRAG)
• Agents & RAG: Agent harnessing (routing, tool-calling, guardrails, multi-agent orchestration); CrewAI, LangChain, LlamaIndex; RAG & GraphRAG

• Detection & Segmentation: YOLO (v3–v11), RT-DETR, Mask R-CNN, UNet, nnU-Net
• Foundation Models: SAM/SAM2, DINOv1–v3, CLIP, SigLIP, ViT
• Libraries: Ultralytics, OpenCV, SAHI, Tesseract / PaddleOCR

• On-premise GPU inference infrastructure, model serving, and deployment strategy
• Edge & Optimization: NVIDIA Jetson, MIG, ONNX

• 3D Processing: MONAI, VTK, ITK, 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 sensing.
• Outcome: Achieved 99.3% accuracy in gait recognition on the Widar 3.0 dataset.
• Impact: Built real-time intrusion detection and indoor localization systems using Deep Learning and Flutter.

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: Most of my work involves proprietary software for enterprise, medical, and government AI, so it is not publicly available. My public repositories showcase research experiments, open-source contributions, and educational tutorials.

Selected Work (high-level):

• AI Recruitment Platform: End-to-end hiring pipeline — from CV ingestion to AI-assisted video-interview analysis and evidence-based candidate comparison.
• Voice-to-Voice Screening Agent: Near-real-time interview agent with a routed multi-agent design and guardrails, built on self-hosted LLMs served on H200 GPUs.
• Medical 3D Segmentation: Production APIs for multi-organ and tumor segmentation with optimized low-latency inference.
• EraBot: Agentic RAG + VLM assistant for clinical/radiology workflows.
• AutoML Vision Platform: Automated computer-vision model training and annotation tooling.

🤝 Let's Connect!

Open to collaborations on Medical AI, LLMs, Computer Vision, and AI Platforms.