System design · Free PDF
AI System Design
A framework for the design round and worked cases: support bot, document search, coding agent and more.
System Design for RAG and LLM Apps
Ingestion, chunking, retrieval and reranking, grounding, and how you evaluate the answers.
Free PDF · sent by email the day it’s ready
Who it’s for
For engineers asked to design search, Q&A or assistant features over a company’s own documents.
Expect follow-ups on retrieval quality, freshness, access control and evaluation.
What it covers
- Ingestion
- Retrieval
- Grounding
- Evaluation
A question from the guide
One of the 16 questions, with its model answer.
RAG design decisions
When would you not use RAG?
Model answer
When the knowledge isn't unstructured text that people search by meaning. If the corpus is small and stable, I paste it into the prompt: no retrieval step means no retrieval misses. If the question is structured, like "orders over $500 last month," that's a SQL query behind a tool, because embeddings are fuzzy and the question is exact. If the data changes per request, like an account balance or the user's current plan, I call the live API instead of indexing a snapshot that goes stale. And if it's general, stable knowledge the model already handles well, retrieval only adds latency.
My two sorting questions are: is it structured, and does it change per request? A yes to either usually means a tool, not an index. RAG's lane is semi-stable documents people search by meaning, like a handbook or a help center.
- What they’re checking:
- Whether you treat RAG as one option among several and route knowledge by its shape, instead of reaching for a vector database by default.
- Likely follow-up:
- A user asks "what plan am I on?" How would you answer it, and why not from the index?
What the guide will include
A framework for the round
Requirements, architecture, tradeoffs, evaluation and failure modes, in the order you say them.
Worked designs
Prompts worked end to end the way you would answer them at the whiteboard, with the diagram.
Tradeoffs to name
The choices interviewers push on, and what each option costs in latency, money and risk.
Failure modes and evals
How the system breaks in production, how you would notice, and what you would measure.
A PDF you can print
Read it on your phone before the call, or print the sections you want to drill.
Planned contents
The guide isn’t out yet, so chapters and their order may still change.
- 1. How the design round runs
- 2. Ingestion
- 3. Retrieval
- 4. Grounding
- 5. Evaluation
- 6. Practice prompts
Free PDF
All 16 questions with model answers
A model answer for every question, what the interviewer is checking, the follow-up and the mistake to avoid.
Questions about the guide
Is the PDF really free?
Yes. Leave your email and we send the PDF the day it is ready. No payment details, and every email has a one-click unsubscribe link.
How is it different from the questions page?
The questions page lists every question with a handful of sample answers. The PDF has a model answer for each question, plus what the interviewer is checking, the likely follow-up and the mistake to avoid.
Is it based on a specific company’s interviews?
No. It is written from the concepts the role or system depends on. We don’t name companies or claim to know what any one company asks.
When will it be ready?
It isn’t out yet. Leave your email and we send it the day it is; nothing else arrives before then.
Related guides
System design · Free PDF
System Design for AI Agents
Agent loops, tools, state and memory, permissions, and how agents fail in production.
System design · Free PDF
System Design for Databases
Choosing a database, indexing, replication, sharding, consistency and vector stores.