The Compute Blog

Writing

I write about how modern AI systems actually work: the tools they use, the knowledge they retrieve, and the engineering that makes them reliable. Start with the reading path below, or catch up on recent posts.

Full articles live on The Compute Blog. This page is a guide to where to begin.

A curated learning path

Start here

Read in order: connect an agent to tools, give it useful context, evaluate its behavior, then decide when to bring in more agents.

  1. Fundamentals · Basics

    What are AI agents?

    What really makes an AI agent?

  2. Tooling · Model Context Protocol

    What is MCP

    Start with how agents connect to tools and data: MCP hosts, clients, servers, and the capabilities they exchange.

  3. Retrieval · RAG architecture

    Retrieval Augment Generation

    What is a RAG system and how does it work?

  4. Retrieval · RAG architecture

    RAG Architect’s Handbook

    Move beyond a basic retrieval pipeline with query transformation, reranking, GraphRAG, and adaptive patterns—and the trade-offs behind them.

  5. Evaluation · Agent reliability

    Evaluating Agentic Systems

    Check whether an agent actually works by evaluating its tool calls and final answers, using golden datasets and a practical DeepEval example.

  6. Orchestration · Multi-agent systems

    Multi-Agent Systems

    Decide when multiple agents are worth the coordination cost, then explore orchestration patterns with Microsoft Agent Framework and Google ADK.

From the feed

Recent posts

Up to 20 recent entries from the blog’s RSS feed, refreshed when this site is built. For older posts and anything published since, visit the full blog.