Bharatsinh Raj·Oct 28How to Evaluate LLM Applications: A Practical Guide to LLM Evals for DevelopersHere’s an uncomfortable question: how do you know your LLM application got better after your last change? (Disclosure: this story was…
Bharatsinh Raj·Oct 28How to Reduce LLM API Costs: 8 Proven Ways to Cut Your OpenAI and Claude BillThe first month of an LLM project is usually cheap. A handful of testers, a few thousand tokens a day, and a bill that rounds to lunch…
Bharatsinh Raj·Oct 28Context Engineering vs Prompt Engineering: The Skill That Matters Now for LLM DevelopersA few years ago, “prompt engineering” meant finding the magic words. Add “think step by step.” Promise the model a tip. Pretend it’s an…
Bharatsinh Raj·Oct 28Model Context Protocol (MCP) Explained: What Every AI Developer Should KnowIf you’ve built anything with LLM tool calling, you know the pain. Every integration is custom. Slack gets one wrapper, your database gets…
Bharatsinh Raj·Oct 28Building AI Agents for Production: 7 Hard Lessons on Agentic AI That Actually WorksEvery AI agent demo looks great. The agent browses, calls three tools, writes a summary, and everyone claps. Then you ship it, and by…
Bharatsinh Raj·Oct 7Agent Harness Engineering: Why the Code Around Your LLM Matters More Than the ModelA recent engineering write-up on running a multi-agent pipeline in production made a point I keep coming back to: the model is close to…
Bharatsinh Raj·Oct 7AI Coding Agents Are Burning Budgets: A Practical Guide to LLM Cost ControlEveryone loved AI coding agents when the bill was a rounding error. Then the agents got better, people let them run longer, and finance…
Bharatsinh Raj·Oct 7RAG Chunking Mistakes That Quietly Break Retrieval (and How to Fix Them)Most RAG failures I read about start before the model sees a single token. The right answer is sitting in the corpus, the vector database…
Bharatsinh Raj·Oct 7Context Engineering for AI Coding Agents: What a Four-Rule CLAUDE.md Teaches UsEarlier this year a single markdown file topped GitHub’s trending list. No code, no dependencies, no model. One article put it at roughly…
Bharatsinh Raj·Oct 7Agent Memory in Production: What to Store, What to Forget, and How to Avoid Stale FactsAsk any team that has worked together for a while and you will hear the same questions: who made that decision, why did we pick this…
