Design an AI gateway for every LLM call across a company. What must it control?
A strong gateway centralizes model routing, provider failover, rate limits, cost tracking, audit logs, PII controls, prompt/version...
Practice AI interview questionsuntil they stick.
2,049 questions across 28 AI topics. Study expert answers, quiz yourself, and let spaced repetition bring weak questions back.
ML fundamentalsLinear algebraOptimizationRLTransformersRAGAgentsEvaluationLLMOpsSystem design
Free forever. No credit card. Built for AI interviews.
14 real questions rotating from the bank
Your AI agent deleted a production database. How do you prevent irreversible actions?
This should never have been possible in the first place. The fix is layered prevention, not just hoping the LLM makes good choices....
Your RAG system is hallucinating despite having the right context. How do you fix it?
The retrieval is correct but the LLM is ignoring or misinterpreting the context. This is a generation problem, not a retrieval probl...
One LLM provider outage took down your product. How do you design failover?
Remove provider lock-in behind a gateway, classify requests by quality/cost tolerance, keep compatible fallback models warm, test pr...
2,049questions
28topics
4practice modes
1study loop
One calm loop for interview prep.
The product is built for repeated use: read, recall, review, and move on without fighting the interface.
Read one category deeply.
Browse the question bank by topic, open a question, and reveal only when you are ready.
Force recall before the reveal.
Pick topics, answer from memory, then mark what you knew and what needs work.
Let weak questions return.
Rate answers from Again to Easy. ACE AI keeps the queue honest without extra setup.
Know what is actually improving.
Track confidence, mastery, streaks, and coverage without turning prep into admin.
Flip a real question card.
Reveal the answer, rate how well you knew it, and let weak cards come back sooner. This is the core study loop without the dashboard noise.
Reveal answerRate recallReturn when due
Your AI agent deleted a production database. How do you prevent irreversible actions?This should never have been possible in the first place. The fix is layered prevention, not just hoping the LLM makes good choices. Layer 1: permission scoping. The agent's database credentials should have read-only access by default. No DELETE, DROP, UPDATE, or TRUNCATE permissi...
Review less. Remember more.
Missed cards return soon. Known cards wait. The schedule stays out of the way until it has something useful to ask.
A single read fades fast. Timed review keeps recall high without forcing you to manage the calendar.
47
Every cell is a day. You can see whether the prep is actually happening before the interview calendar gets loud.
From fundamentals to production judgment.
Categories are grouped by how interview loops usually progress, so it is easy to decide what to study next.
01
Fundamentals
Math, ML, optimization, RL, deep learning, transformers, and LLM internals.
02
Core Techniques
Prompting, retrieval, embeddings, fine-tuning, multimodal work, and applied model behavior.
03
Building Systems
Agent workflows, implementation judgment, evaluation loops, and shipping habits.
04
Production
Serving, infrastructure, system design, and safety work that keeps AI systems alive.
05
Leadership & Strategy
Product tradeoffs, strategy judgment, and behavioral loops for senior interviews.
Open the dashboard and start with one question.
The fastest path is simple: pick a topic, answer out loud, reveal, rate, and keep moving.