Mohamed A.

🚀 I build reliable LLM, RAG, and AI agent systems that go beyond demo chatbots — with retrieval, evaluation, guardrails, tool use, and production-ready engineering. My work includes multilingual RAG, document intelligence, AI agents, OCR + multimodal systems, and computer vision. 🏆 SELECTED RESULTS • Improved an object detection system from 84.5% → 92.0% mAP@50 • Built and evaluated multilingual RAG pipelines using Recall@K, MRR and nDCG • Built a multimodal OCR pipeline that outperformed Azure Document Intelligence while reducing processing cost • Built an AI legal-compliance system producing traceable PASS / FAIL / REVIEW decisions with evidence and human review 🧠 LLMs, RAG & AI AGENTS I can help you build: ✓ RAG systems over internal documents and knowledge bases ✓ AI agents that call APIs, tools and business systems ✓ Document extraction and document-understanding pipelines ✓ AI assistants with citations, validation and audit trails ✓ LLM/RAG evaluation and hallucination testing ✓ Embedding, retrieval and reranking optimization ✓ Multilingual AI systems ✓ OCR + LLM / multimodal pipelines 👁️ COMPUTER VISION I also work on: ✓ Object detection and segmentation ✓ Multimodal vision models ✓ Model optimization and inference pipelines ✓ PyTorch and Triton deployment 🏢 EXPERIENCE At Linde, I work on document intelligence and AI systems, including multilingual retrieval, embedding and reranking evaluation, OCR pipelines and multimodal document understanding. My research background is in reliable AI. I have published peer-reviewed research in IEEE Access on adversarial robustness and study Data Science at LMU Munich as part of the relAI — Konrad Zuse School of Excellence in Reliable AI. ⚙️ HOW I WORK I don't just connect an LLM API and call it an AI product. I focus on: ✓ measurable evaluation ✓ reliable outputs ✓ traceability ✓ latency and cost ✓ maintainable production architecture If you're building an AI product — or already have one that is hallucinating, unreliable, expensive, or difficult to scale — send me a message and tell me what you're trying to build.