Test yourself the way the loop will.
Pick a topic or take the whole spread, choose how hard and how long, and answer real multiple-choice scenarios drawn from what AI infra interviews actually probe. Every answer comes back with the why, and with a link into the worked material that goes deeper. The questions rotate between takes, so a retake is a real re-measure, not a memory test.
What a test looks like
Two samples from the bank, answer and explanation included. Inside a test, the explanation appears after you commit to an answer, which is where the learning happens.
AI Infrastructure System Design · easy · sampleWhich seven boxes make up the reference shape of an LLM inference platform?Show the options and the answer →
The gateway authenticates, rate-limits and validates; the router picks a replica by prefix locality and load; the engine's scheduler batches and admits requests; replicas hold model copies across one or more GPUs; the KV tier is the paged pool on GPU with optional offload; the autoscaler sizes the pool with cold starts in mind; observability tracks TTFT, TPOT, goodput and KV utilization. A design answer names each box's job and its failure mode, then sizes the pool with a stated chain of arithmetic.
AI Infrastructure System Design · easy · sampleWhy is a least-connections load balancer the wrong router for LLM replicas?Show the options and the answer →
A 300-token completion and a 30,000-token summary are one connection each, but differ by two orders of magnitude in prefill work and KV footprint; and a replica that already holds a conversation's prefix answers in 12 ms of prefill where a cold one needs 200 ms. A cost-aware router scores each replica as expected prefill after cache hits plus expected queue wait, reading pending tokens, running sequences and KV occupancy from the engines, and picks the minimum; near saturation the queue term overrides affinity.
Eleven topics, three difficulties
Easy questions teach the vocabulary. Medium is applied judgment. Hard is where staff-level tradeoffs and 3am triage live. Every topic maps to a question journey and a concept track, so a weak score has an obvious next step.
Where practice tests fit
Before: find your gaps
The free readiness assessment diagnoses which of the 11 areas you have actually covered and builds a study plan from the misses.
Take the assessment →During: drill and measure
Practice tests are the reps. Learn mode explains every answer as you go; exam mode holds the reveal for the end. The score is a recall check, honestly framed: it measures what you can recognize under time, not whether you can run a discovery call.
After: go deep where you missed
Every explanation links the concept or worked question behind it: 413 full questions with worked answers, 150 concepts, three courses.
Browse the full questions →Common questions
- Are the practice tests free?
- Yes, with a free account: easy-difficulty tests up to 20 questions, on any topic, with full explanations after every question. Medium, hard and adaptive tests, and longer runs up to 100 questions, are part of premium.
- Do the questions repeat if I retake a test?
- The bank rotates: a retake deals questions you have not seen before, until you have genuinely exhausted the pool for that topic and difficulty. The bank currently holds 400 questions and grows.
- How is this different from the readiness assessment?
- The assessment diagnoses which of the 11 areas you have covered and builds a study plan. Practice tests are the drilling half: score yourself on a topic, learn from every explanation, and watch the score move as you study.
- What do the tests cover?
- The same 11 areas real AI infra loops test: RAG and agents, LLM fundamentals, system design, SQL and data engineering, MLOps, ML infrastructure and GPUs, classical ML, ML product design, coding, AI security, and the customer-facing rounds.