How This Course Connects to the GenAI Lab

Every chapter from here forward ends with a Lab Bridge, a pointer to one or more real, gradable levels in the GenAI Lab that put what you just read into practice. This chapter explains exactly what you will find when you get there, so the first level does not feel unfamiliar.

The GenAI Lab is not a text box where you type a prompt and get a grade back. It is a visual canvas, the same spirit as the HLD Lab's architecture canvas, but built for LLM pipelines instead of web infrastructure. Every level opens with a case brief from a fictional client at "the Agency": a real-sounding problem, a budget, a latency SLA, and a set of constraints. Your job is to drag the right components onto the canvas, wire them together in the right order, and configure them correctly.

• -PromptTemplate: structures the system and user message that actually gets sent to the model.
• -LLM: the model call itself, configurable for things like max_tokens, temperature, and which underlying model to use.
• -ConversationMemory: carries history between turns, and can be configured to summarize after a certain number of turns.
• -InferenceRouter: sends a request to a cheaper or more expensive model depending on complexity.
• -SemanticCache, RateLimiter, StreamHandler, InputFilter, OutputGuard, AgentLoop, ToolRegistry: each one exists because a specific real failure mode needs it, and you will meet every single one of these by name across this phase.

Behind the scenes, every level has a list of missions, and each mission checks something specific and mechanical: is a required component present on the canvas, is it connected to the right neighbor in the right direction, does a specific configuration field satisfy a constraint (max_tokens at or under a limit, a cache TTL within a range, a temperature set low enough for a deterministic use case). This is not a fuzzy, subjective grade. It is closer to how the HLD Lab checks your architecture: concrete, inspectable, and the same every time you submit.

Before the 50 graded levels of Acts 1 through 5, eight tutorial levels (numbered -8 through -1) walk you through the canvas mechanics themselves with no scoring pressure: meeting the model, giving it instructions, giving it memory, understanding embeddings, chunking, guardrails, cost, and assembling your very first end-to-end pipeline. If you have never touched the GenAI Lab before, start there right now, before reading another word of this course. It takes about twenty minutes and everything after this chapter will make more sense with the canvas already in your hands.