AI ENGINEERING
AI Engineering
How LLM-powered software is built: agents, tool calling, retrieval, memory, guardrails and evals. Each episode is an animated walkthrough with notes and interview questions, the same way the System Design track works.
All 5 are live. Each one is an animated walkthrough you can step through, written notes, and the interview questions with answers.
Agents
What an agent is, how it calls tools, loops, remembers, retrieves and stays safe.
- 1 What is an AI agentWhat separates an AI agent from a chatbot: a language model that chooses actions, tools that touch the outside world, memory that carries state, and a loop that runs until the goal is met. Workflows versus agents, a worked trace, when an agent is worth it, how errors compound across steps, and the cost numbers to quote.
- 2 Tool callingHow a language model calls functions it cannot run itself: tool definitions with names, descriptions and JSON schemas, the request and result round trip, parallel calls, validating arguments and permissions, returning errors the model can fix, guarding write actions, designing tools a model uses well, forced tool use for structured output, MCP, and injection through tool results.
- 3 The agent loopThe control loop at the heart of every agent: reason, act, observe and repeat. Stopping rules and budgets, a step-by-step trace of a coding agent, why context grows and how to trim, summarise and offload it, planning and self-checks, catching stuck loops, sub-agents, pausing for people, checkpoints for long runs, and tracing every step.
- 4 Retrieval-augmented generation (RAG)How retrieval-augmented generation grounds a model in your own documents: ingesting and chunking, embeddings and vector search, hybrid keyword plus vector retrieval, reranking, building a prompt with citations, rewriting queries and retrieval as an agent tool, freshness and permissions, measuring retrieval and faithfulness, common failures, and when to pick RAG over long context or fine-tuning.
- 5 Guardrails and evalsHow production AI systems stay safe and get measurably better: layered guardrails on input, tools and output, prompt injection and least privilege, human approval, then evaluation: building a task set from real traces, code graders, model judges and people, judging whole agent runs, gating releases on regressions, and monitoring quality, cost and drift in production.
Keep going
System Design covers the servers, data stores and queues these agents run on. DSA keeps your coding sharp for the interview.