markdown

I am Muhammad Fayaz Khan, an LLM Engineer and ML practitioner from Pakistan. I first mastered data communication using Pandas and NumPy. Later, I specialized in predictive modeling and machine learning with Scikit-Learn then Deep artificial neural networks and deep learning with TensorFlow and Pyrorch. Currently, I focus on developing enterprise-grade RAG systems and Agentic AI workflows using FastAPI, LangChain, and Vector Databases. I focus primarily on maximizing retrieval accuracy, implementing latency optimization, and achieving LLM cost reduction for production-ready AI applications. As a self-taught professional, I am executing an advanced LLM Engineering roadmap in 2026.

Enterprise FAQ agent designed to ingest business PDFs and answer domain-specific queries using Hybrid Search and Cross-Encoder Reranking workflows. The architecture is built using FastAPI, ChromaDB, LangChain, Pydantic, Chainlit, and Docker. It is engineered for real client deployment featuring persistent chat history and exhaustive RAGAS evaluation.

Production-grade RAG system utilizing Recursive Semantic Chunking and Cross-Encoder Reranking strategies to query private PDF data safely. The pipeline achieved high faithfulness and answer relevancy scores validated via the RAGAS evaluation framework. Technical stack includes FastAPI, ChromaDB, LangChain, and RAGAS.

Autonomous agent built with structured Tool Calling mechanics and validation pipelines producing precise JSON outputs via Pydantic v2. Developed to handle complex, multi-step business workflows autonomously, engineered using LangChain, and optimized thoroughly for multi-agent cooperation systems.

Predictive model engineered for the Rossmann retail chain to execute highly accurate sales forecasting tasks. Applied advanced feature engineering strategies along with ensemble methods to achieve competitive accuracy metrics on time-series retail datasets. Technical stack utilizes Python, Scikit-Learn, and Pandas.

Machine learning classification model constructed to predict customer churn trends. The codebase contains full end-to-end data preprocessing pipelines, strategic feature selection setups, and extensive model evaluation stages using precision-recall analysis. Technical stack includes Python, Scikit-Learn, Pandas, and FastAPI.

Classification model deployed successfully as an interactive Streamlit web application to provide instant diabetes risk prediction based on complex patient health parameters. Technical stack utilizes Python, Scikit-Learn, and Streamlit.

End-to-end regression pipeline developed for predicting real estate house pricing trends. The architecture incorporates advanced data cleaning, exploratory data analysis, and multi-model comparison matrices. Technical stack includes Python, Scikit-Learn, and Streamlit.

Deep learning image classification model utilizing Convolutional Neural Networks built entirely with PyTorch. The model is deployed as an interactive user interface via Streamlit. Technical stack utilizes PyTorch, custom CNN layers, and Streamlit.

• Optimizing LLM API costs and execution latency metrics for production-ready client applications
• Implementing production LLM Caching mechanisms via GPTCache, reducing operational API costs by up to 40 percent
• Engineering stateful, cyclical multi-agent orchestration architectures and workflows utilizing LangGraph
• Designing robust RAG 2.0 pipelines featuring Hybrid Search architectures and Neural Reranking systems for enterprise environments
• Interfacing and integrating multiple core LLM providers including OpenAI, Anthropic, Gemini, and Groq
• Executing red-teaming methodologies and evaluating generative LLM outputs using RAGAS and LLM-as-Judge frameworks
• Formulating advanced Prompt Engineering designs including Zero-shot, Few-shot, Chain-of-Thought, ReAct Patterns, and Context Engineering
• Developing Agentic AI frameworks utilizing LangGraph, CrewAI, LangChain, Tool Calling methodologies, and Multi-Agent Workflows
• Structuring backend services and APIs using FastAPI, asynchronous Python, decorators, generators, Pydantic v2, and the Instructor library
• Managing vector databases effectively including production setups with ChromaDB, Qdrant, and Pinecone
• Applying foundational AI and ML principles including Machine Learning, Deep Learning, PyTorch, CNN architectures, and Scikit-Learn pipelines
• Implementing DevOps tooling setups including Docker, Docker Compose, Git, GitHub version control, and Streamlit deployment

• 12th Grade, Intermediate Computer Science, Completed
• Self-Taught LLM Engineering Curriculum, 2026

• Lakki Marwat, KPK, Pakistan, also operating within the Rawalpindi and Islamabad region

• Email: muhammadfayazkhan50@gmail.com
• Phone: 03120770266
• LinkedIn: https://linkedin.com/in/muhammad-fayaz-khan-271487381
• GitHub: https://github.com/mfayazkhan50-AI
• TikTok: https://tiktok.com/@aiwithmfk