Engineering Notes
Thoughts and Ideas on AI by Muthukrishnan
Tag: Llms
RLHF and Preference Learning Teaching Agents What Humans Actually Want
Master reinforcement learning from human feedback — the algorithm behind ChatGPT and modern aligned agents — from reward modeling and PPO to Direct Preference Optimization.Systems That Search for New Algorithms Automatically
An introduction to AlphaEvolve, DeepMind's new AI agent that uses evolutionary search and LLM-driven code synthesis to autonomously discover and optimize algorithms, outperforming human-designed solutions in several domains.TDD with Coding Agents to Build a Rate Limiting Service
## Problem Overview We'll build a sophisticated rate limiting service that supports: - Multiple rate limiting algorith...The Complete Claude Code Best Practices Guide
*A Comprehensive Manual Compiled from Developer Experiences and Community Knowledge* ## Table of Contents 1. [Introduc...Why MMLU Matters and What Massive Multitask Language Understanding Really Tests
A personal dive into MMLU - the benchmark that's reshaping how we measure AI intelligence across dozens of academic subjects.Understanding Mixture of Experts from First Principles
One of the most exciting architectural innovations in recent years is the **Mixture of Experts (MoE)**. It's a key reas...AI Agent Communication Protocols MCP, ACP, A2A, and ANP
An overview of the most important communication protocols for multi-agent systems, and how they enable interoperability and scalability.Small Language Models are the Future of Agentic AI
I recently came across a paper titled *"Small Language Models are the Future of Agentic AI,"* and it got me thinking. T...Meet AlphaEvolve the AI That Discovers New Algorithms
An introduction to AlphaEvolve, DeepMind's new AI agent that uses evolutionary search and LLM-driven code synthesis to autonomously discover and optimize algorithms, outperforming human-designed solutions in several domains.Learning CrewAI in a Step-by-Step Guide with 10 Concepts
A beginner-friendly guide to mastering CrewAI, from basic agents and tasks to advanced multi-agent workflows with tools and custom LLMs.An Introduction to Context Engineering for AI Code Generation
How to prevent AI coding assistants from generating outdated or deprecated code by using real-time context engineering.A Developers Guide to Effective AI Coding Agents
Moving beyond basic code completion to a structured, developer-led workflow that harnesses the full potential of AI coding agents.Deep Research Applications of Modern LLMs Generated through Deep Research
Exploring how modern AI systems are evolving beyond conversational chatbots to become autonomous research agents capable of conducting comprehensive, PhD-level investigations.