Ahmed D.

I build production RAG and LLM agent systems that actually ship, not notebook demos. Recently deployed a multi-agent RAG platform serving 500+ daily users at 92% query accuracy on Pinecone, LangGraph, and AWS. I'm an ML engineer with 5+ years building and deploying AI systems for global clients, specialized in generative AI, retrieval-augmented generation, and production-scale ML pipelines. I've shipped end-to-end platforms across LLM agents, computer vision, and MLOps from data pipeline to deployed, observable, low-latency service. What I do best: • RAG & LLM systems — retrieval pipelines, vector search (Pinecone, pgvector), multi-agent orchestration with LangGraph, WhatsApp/CRM integrations • MLOps & deployment — Docker, Kubernetes (EKS), AWS (Lambda, S3, SageMaker, EC2), CI/CD, 99.5% uptime services • Inference optimization — quantization (INT8/FP16), TensorRT, ONNX, batching; cut LLM latency 60% and boosted inference speed 30% on real projects • Computer vision — action recognition, medical image segmentation, image generation, edge deployment Recent results: • Production RAG assistant — 500+ daily users, 92% accuracy, 60% lower latency • Medical imaging — +15% diagnostic accuracy across 100K+ images; 94% segmentation accuracy on hemorrhage detection • Edge AI — real-time shoplifting detection running on-device with no cloud dependency Stack: Python, PyTorch, TensorFlow, LangChain, LangGraph, Hugging Face, LLMs, RAG, Stable Diffusion, AWS, Kubernetes, Docker, Pinecone, PostgreSQL, TensorRT, ONNX. If your AI project needs to work in production reliably, fast, and observable let's talk. Send a note about what you're building and I'll tell you exactly how I'd approach it.