Learning to pass
the interview
Two interview tracks in one place: software engineering and AI engineering. SWE covers data structures, algorithms, and system design, explained visually without the math wall. The AI track starts from zero: if you have used ChatGPT but never built the systems behind it, it takes you from a blank chat window to interview-ready. Interactive walkthroughs, a Python playground, and progress tracking on both.
AI engineering
9 topicsThe second track. You have used ChatGPT; this is what the systems behind it actually are, why they fail, and how you answer for that in interviews. Built to start from zero - no AI background assumed.
Semantic Search & Document QA
Score a document against a question, rank the scores, and fit what you keep in the space a model can read. Three pieces of plain math, built from scratch.
Prompting & Context Windows
The words you send decide the words you get back, and they all fit in a fixed space. Learn to change a prompt on purpose, not on vibes.
RAG System Design in Production
Answer questions from documents the system looks up mid-conversation, then break the pipeline and diagnose exactly which part failed.
Fine-Tuning: When, and Why
Some failures a better prompt fixes, some only training fixes. Learn the cost ladder and the decision factors interviewers ask for.
LLM Evals & Judge Calibration
How you prove a prompt or model change actually got better, with measured tests instead of eyeballing outputs.
AI Agents: Loops, Tools, and Stopping
Give a model tools and it will loop, stall, or wander. Learn the agentic loop and the points where you decide it should stop.
Inference Economics & Latency
Why one AI request costs pennies and another costs dollars, and what moves the wait time between them.
Guardrails & Prompt Injection
Untrusted text joins the prompt from more doors than the user. Learn where it gets blocked and how injection attacks run.
Serving AI Reliably
The model behind your feature can change under you at any time. Learn what to watch, what to fall back to, and when to roll back.