MCP vs Tool Calling vs Skills — And Where RAG Fits In
The three ways to extend an LLM aren't interchangeable — each solves a different problem at a different layer. Here's the mental model, with RAG as a fourth...
Writing
Technical articles and insights on AI, MLOps, and production machine learning.
Technical deep dives and practical guides on agentic AI, RAG evaluation, LangGraph, MLOps, and deploying AI systems in healthcare and industrial settings.
Building a Conversational Voice Agent for IMDB Data with LangGraph, FAISS, and Streamlit
A deep dive into the architecture and design decisions behind a GenAI-powered movie assistant that combines structured pandas queries, semantic vector search, and LLM reasoning into a single...
Retrieval-Augmented Generation: A Comprehensive Technical Guide
A deep dive into RAG architecture, embeddings, chunking strategies, retrieval patterns, and production best practices for building grounded, hallucination-resistant AI systems.
Why Most Agentic AI Systems Fail in Production (And How to Fix That)
Agentic AI demos look great. Production deployments rarely do. Here's the engineering discipline that bridges the gap.
RAG Retrieval Metrics You Should Actually Be Tracking
Hit Rate@K, MRR, NDCG — what they measure, when each one matters, and how to implement them for your RAG system.
MLOps Monitoring: The Tools Actually Worth Using in 2026
A practitioner's comparison of Evidently, Arize Phoenix, Azure ML Monitor, and LangSmith for production ML and LLM monitoring — based on real deployments.
Ai Human Side 2025
The Human Side of AI: What the Data Reveals — and What Businesses Are Getting Wrong *By Mohcine Madkour, PhD Senior AI/ML Engineer & Architect* The 2025 AI...
Agentic AI in Production: What No One Tells You↗
The engineering disciplines that separate agentic AI demos from reliable production systems — evaluation, monitoring, and CI/CD quality gates.
HL7 FHIR Integration for Clinical AI: A Practitioner's Guide
How to connect clinical AI models to EHR systems using HL7 v2 and FHIR R4, based on lessons from MySurgeryRisk at UF Health.
Real-Time Multi-Object Tracking with YOLOv8 and ByteTrack
Building a production warehouse computer vision system — object detection, persistent tracking, and QR-code identity resolution at 30fps.