Muhammad M.

Overview: I build AI systems that work on real data and not just demos. As a Data Scientist and Computer Vision Engineer, I’ve taken projects from customer requirements through to deployment and validation across industrial, medical, and LLM-based applications. What I’ve worked on: Industrial datasets & visual inspection: I’ve delivered AI inspection solutions across five enterprise manufacturing projects, fine-tuning YOLO, RF-DETR, and SAM models on 15,000+ images spanning 20+ defect classes. Sensor data: I’ve built pipelines that combine RGB and thermal sensor evidence from robotic inspection runs into structured, machine-readable reports. Medical imaging: I build deep learning pipelines for medical image analysis, with hands-on experience in cardiac MRI segmentation using PyTorch, MONAI, and Vision Transformer and Attention U-Net architectures. I benchmark models for accuracy, speed, and deployment feasibility, then ship them as Dockerized FastAPI services with MLflow tracking. My cardiac MRI work reached 81.73% IoU and cut a manual 10-minute workflow to under one minute. RAG chatbots & retrieval pipelines: I build LLM assistants grounded in your data using LangChain, LlamaIndex, FAISS, and Elasticsearch, with a focus on accurate, verifiable answers. Agentic workflows & automation: I design tool-calling agents and automate workflows with n8n and Python, including an LLM-powered engineering assistant that saved 15-20 minutes on routine tasks. Tech stack: Python, PyTorch, TensorFlow, Hugging Face, OpenCV, LangChain, LangGraph, FastAPI, PostgreSQL, Docker, Azure, MLflow, n8n How I work: I ask the right questions up front, communicate clearly, and test against your acceptance criteria before handing anything over. If you have messy data, a vision problem, or an AI workflow you want to automate, send me a message and let’s talk.