Know what they actually ask, before you walk in.
Indwar collects and verifies technical interview questions reported by real candidates, then pairs them with structured tracks so you learn the underlying material instead of memorising answers.
Free, and no account needed for any of it.
Collections
Start from the loop, not from a topic list
Each collection is a shortlist of what that company has been asking recently, with the round it showed up in and how confident we are.
Anthropic
24Heavy emphasis on practical coding under time pressure and on reasoning out loud about safety trade-offs.
OpenAI
41Applied ML systems, inference-time performance work, and a debugging round on unfamiliar code.
Stripe
33Integration-style problems on a real API surface, plus correctness questions around money and idempotency.
Databricks
29Distributed data processing, query planning intuition, and one deep round on a system you have actually operated.
Snowflake
26Storage and concurrency internals, with a strong preference for candidates who can reason about cost.
Uber
38Geospatial and real-time matching problems, then a design round on something with a live dispatch loop.
Not seeing your company? Their loops vary so much by team and org that a single shortlist would mislead more than it helps. We would rather ship nothing than ship a list you cannot trust. Submit what you were asked and it moves up the queue.
Learning tracks
Learn the material, not the answer key
A question bank on its own teaches you to recognise problems you have seen. The tracks are there so you can solve the ones you have not.
Core coding patterns
The small set of shapes that most coding rounds reduce to. Learn the shape, then the variations, then the edge cases interviewers actually probe.
24 lessons
System design
Work through designs the way a staff engineer would: constraints first, then the boring parts, then the one hard trade-off the interviewer cares about.
21 lessons
ML systems
For infra and applied-ML loops: how models get served, evaluated, and rolled back when they regress in production.
16 lessons
Debugging unfamiliar code
An increasingly common round: you are dropped into a repo you have never seen and asked to fix a failing test in 45 minutes.
11 lessons
Behavioural and scope
Structured stories for impact, conflict, and failure — plus how to answer scope questions without overclaiming.
9 lessons
How to use it
Three weeks, used well
Most people over-study breadth and under-study the two rounds they are most likely to fail. This is the order we suggest instead.
Pick the loop you are in
Start from a company collection so you know what shape of question to expect before you spend a single evening studying.
Fill the gaps with a track
Each question links to the lessons behind it. When you miss one, you learn the underlying idea instead of memorising an answer.
Practise under real constraints
Timed attempts, no hints on the first pass, and a written post-mortem for every problem you did not finish cleanly.
Why Indwar
A small set that is right beats a large set that is stale
Reported, then corroborated
A question only enters a collection once separate candidates describe the same problem independently. Single reports sit in a holding queue.
Scoped to one loop
Collections are per company, not per topic. What you read is what that company has actually been asking recently.
Aged out on purpose
Anything we cannot re-corroborate within a year moves to an archive. Old sets shrink instead of quietly growing wrong.
FAQ
The obvious questions
Where do the questions come from?
Engineers who have recently interviewed submit what they were asked. Nothing under NDA, no take-home content, no proprietary code — just the shape of the problem and the round it appeared in.
How is a question verified?
We wait for independent reports from candidates who did not interview together, then reconcile the versions into one description. If the accounts conflict, the question stays unpublished.
Why is my company missing?
Either we do not have enough corroborated reports yet, or the loop varies too much between teams to summarise honestly. Meta and Apple fall into the second bucket.
What does it cost?
Nothing. Every collection and every lesson is free right now, with no account and no card. If paid tiers ever arrive, the coding patterns track and a preview of each collection stay free.
Your loop is in two weeks. Start with what they ask.
Every collection and every lesson, free. No account, no card, no trial counting down.