As an AI Engineer, I specialize in Computer Vision, Large Language Models, and RAG systems. Currently at IdaamCloud, I build scalable AI solutions, including real-time voice agents that integrate speech recognition, TTS, and intelligent dialogue[file:1].

Key achievements: Reduced human agent workload by 60% via automated systems, boosted automation efficiency by 30%, cut fraud by 20%, and achieved 95-98% accuracy in CV pipelines[file:1].

Passionate about fine-tuning LLMs, prompt engineering, and production pipelines with LangChain, LangGraph, and PyTorch. Explore my portfolio for demos.

Let's collaborate on innovative AI projects—reach out via LinkedIn!

Specialized in: LangChain • LangGraph • Unsloth • LiveKit • Pipecat • CrewAI • vLLM • LoRA/QLoRA • OpenRLHF • RAG • LLaMA-Factory[file:1]

Specialized in: YOLOv8 • Detectron2 • MediaPipe • MMDetection • Ultralytics • ByteTrack • Vision Transformers • ONNX[file:1]

KEME Brain Health USA - Voice Agent System

• Developed automated call-center assistant reducing human workload by 60% using LiveKit, XTTS API, and Whisper v3[file:1]
• Fine-tuned Qwen3 model with KTO architecture for improved conversational quality[file:1]
• Designed scalable backend with API integrations and timezone detection[file:1]

AI-Supervision System

• Engineered CV-based proctoring platform with YOLO-World for anomaly detection[file:1]
• Built multi-object tracking pipeline (Ultralytics, ByteTrack) for gaze/hand monitoring[file:1]

• Virtual Health Assistant: Increased patient engagement by 30% with LangChain and RAG pipelines[file:1]
• LLM Fraud Detection: Cut fraud losses by 20% using fine-tuned LLaMA-3 and LangGraph[file:1]
• Automation Solutions: Boosted efficiency by 30% with PyTorch and FastAPI[file:1]

• Podcast Insight Agent: LLM-powered analysis engine with RAG, semantic search, and Docker/K8s deployment[file:1]
• Football Tracking System: Tracks players/referees/ball at 95% accuracy using YOLOv8 and ByteTracker[file:1]
• License Plate Recognition: Vehicle detection system with 98% accuracy via YOLOv8, SORT, and EasyOCR[file:1]
• Transformer from Scratch: PyTorch implementation achieving 90% accuracy in NLP/vision tasks[file:1]

• YOLOs-CPP: Contributed to high-performance C++ YOLO inference with ONNX Runtime and OpenCV optimizations.

• Bachelor of Computer Science - Mansoura University (2019-2023)[file:1]

Specialized Training:

• Retrieval-Augmented Generation (RAG) - Abu Bakr Soliman (2024–2025)[file:1]
• Modern Computer Vision with PyTorch (Ongoing)[file:1]
• PyTorch for Deep Learning & ML - freeCodeCamp (2023)[file:1]
• Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow (2023)[file:1]

khaled_expertise = {
    "specialization": ["Computer Vision", "LLMs", "RAG Systems", "Voice Agents"],
    "languages": ["Python", "C++"],
    "frameworks": ["PyTorch", "TensorFlow", "LangChain", "FastAPI"],
    "ai_tools": ["YOLOv8", "Whisper", "GPT", "Transformers"],
    "deployment": ["Docker", "Kubernetes", "AWS", "Linux"],
    "achievements": {
        "workload_reduction": "60%",
        "tracking_accuracy": "95-98%",
        "automation_boost": "30%"
    }
}

⭐️ From Elbhnasy | Building AI Solutions that Matter 🚀