Writing & Insights
Deep dives on Agentic AI systems, LLM architecture, production RAG pipelines, and human-in-the-loop system design.
Why Human-in-the-Loop Architecture is the Creative Core of Agentic AI Systems
In the world of agentic AI and LLM systems, hallucination is the biggest structural challenge. This article explores how human-in-the-loop design becomes the most creative and robust solution in production-grade AI architectures.
Agentic AI Human-in-the-Loop LLM Systems

Read Article →
The Most Important Component of RAG: Building and Evaluating a Retriever
A practical guide to building a baseline retriever for RAG systems. Learn how to implement chunking strategies, generate embeddings, perform vector search using FAISS, and evaluate retrieval performance using Precision@K, Recall@K, and MRR.
RAG Retriever Vector Search LLM Systems

Read Article →
BM25 vs Dense vs Hybrid Retrieval for RAG: Architecture, Trade-offs, and When to Use What
Learn how BM25, Dense Retrieval, and Hybrid Retrieval work in RAG systems, including architecture, ranking strategies, and when to use each approach in production pipelines.
RAG BM25 Dense Retrieval Hybrid Retrieval Vector Search Reranker Information Retrieval

Read Article →
Improving BM25 Retrieval with Query Expansion: A Practical Evaluation Using Precision, Recall, and MRR
An experimental analysis of query expansion techniques for BM25, demonstrating measurable improvements in recall, precision, and MRR.
RAG BM25 Query Expansion LLM Information Retrieval Search Systems

Read Article →
Tree of Thoughts (ToT): Improving Reasoning in RAG and Agentic AI Systems
Improve LLM reasoning using Tree of Thoughts (ToT). Learn why Chain-of-Thought fails, how multi-path reasoning works, and how to implement it in real-world RAG and agentic AI systems.
Tree of Thoughts LLM Reasoning Agentic AI RAG Prompt Engineering Beam Search Multi-Path Reasoning Gemini API

Read Article →
From Research to Revenue: Applying Google’s Small Model Strategy to Build an Intent-Aware Lead Scoring System
Learn how to build an intent-aware lead scoring system using Google’s small model strategy. Combine clickstream data, LLM pipelines, and hybrid system design to prioritize high-conversion leads and drive real business impact.
Lead Scoring Intent Detection Small Models LLM Systems Clickstream Data Agentic AI Human-in-the-Loop AI in Sales

Read Article →
From Query to Action: Inside an MCP-Powered Agentic AI System
Learn how MCP (Model Context Protocol) powers agentic AI systems. Explore architecture, tool orchestration, guardrails, and multi-agent design to build reliable, production-ready AI systems.
MCP Agentic AI LLM Systems AI Architecture Tool Calling Multi-Agent Systems AI Engineering Production AI

Read Article →