Posts

There was a time when learning software development felt like climbing a mountain with bare hands. Many developers from the early web era remember learning HTML, CSS, JavaScript, PHP, and MySQL using nothing more than simple text editors like Notepad. There were no AI copilots, no instant code generation tools, no Stack Overflow answers for every edge case, and no smart IDEs correcting mistakes automatically. We wrote every line manually....

What We Are Building This post explains what a Claude agent is, how it works, and how skills let you write agents that are tiny, flexible, and easy to change - using a real, runnable example from start to finish. The example is a grading agent. By the end, you will have a working agent that: Reads a CSV of student marks Calculates a letter grade (A, B+, C−, etc.) and a GPA (4....

If you’ve heard the buzz about Claude Code but feel a little lost about how it actually works, you’re in the right place. Claude Code isn’t just another AI chatbot stitched into your terminal — it’s a fully configurable development environment where you, the developer, decide how the AI behaves inside your project. But to use it well, you need to understand its anatomy. Don’t worry, it’s simpler than it looks....

From CLAUDE.md to agent teams: a file-by-file breakdown of how a modern Claude Code project is organized in 2026, and a step-by-step plan for adopting it in your own repo.

Introduction Managing user identity across multiple systems is one of the hardest problems in modern software architecture. Especially when those systems belong to different companies. OAuth 2.0 Token Exchange (RFC 8693) is the standardized solution: it lets a central identity provider (IdP) authenticate a user once and issue system-specific access tokens for any downstream application, even across organizational boundaries. In this post, we’ll explore how to build a central IdP system, exchange tokens for different applications, and handle users who may not exist in every system — a problem often seen after a company acquisition....

Introduction Artificial intelligence has come a long way from answering simple questions. Today’s AI agents can browse the web, write and execute code, send emails, manage calendars, and coordinate with other AI agents. All on their own. But how exactly does an AI go from receiving a vague instruction to completing a complex, multi-step task? That’s where the DOE Framework comes in. Standing for Directive, Orchestration, and Execution, this three-layer model describes how modern agentic AI systems are structured, and why that structure matters for building AI that actually works in the real world....

In this post, I will guide you through setting up a local AI chatbot using Ollama, DeepSeek-R1, and Open WebUI. This approach provides flexibility, privacy, and control over your AI interactions without relying on cloud-based services. Also saves cost in the long run. We’ll keep things simple and concise while covering the installation of Ollama, running LLM models like DeepSeek-R1, and using the web interface to interact with them locally....

In this post, I will first define what LlamaIndex, RAG (Retrieval-Augmented Generation) is, describe the basic architecture pipeline that I am building, and implement the concept using Python code. LlamaIndex LlamaIndex is a powerful data framework design to connect custom(text, HTML, pdf, etc.) data sources to the Large Language Models. It acts as an interface to manage the interaction with the LLMs like loading data from source, create the index form the input data, which than used to respond to the users queries....

In this post, I will explain how to implement the Strategy Design Pattern in Node.js. To illustrate this concept, I’ll use the example of integrating multiple payment methods which can be used in the e-commerse and other applications to accept payment from customers. Whether it’s PayPal, bank transfers, or credit card. Strategy Pattern The Strategy Design Pattern is a behavioral design pattern in software engineering that enables the dynamic selection of algorithms at runtime, allowing for greater flexibility and adaptability in the application’s behavior....

In this post, I will create a small Node.js app containing the publisher and message processor, and publish it as an image to Docker Hub. Then, I will use the same image to demonstrate scaling Kubernetes Jobs using KEDA, triggered by Amazon SQS. For demonstration purposes, I will be using a Minikube local Kubernetes cluster, but in reality, this would be done in a production Kubernetes cluster Here is the architectural diagram for our workflow....