Filter by the role, format, or skill you want to sharpen.
Design how file attachments move from upload to a state where AI chat can use them safely, including slow processing and failures.
Seen at xAIProcess cache reads and writes while evicting the least recently used key.
Interview evidence from 3 companies7 short-answer questions on ML fundamentals.
This throttle guards the /v1/messages endpoint. Each account gets a refilling budget measured in units, and bigger generations cost more units. It passes its unit tests and is…
Seen at xAIStrategy-memo interviews at PM-leadership level test whether you can make a decision under ambiguity, with real stakes across safety, compute, revenue, and competitive timing.…
Interview evidence from 2 companiesImplement a K-nearest neighbors classifier from scratch, with a deterministic rule for tied votes.
Seen at xAIThis helper decides whether a customer user can read a file. Paths are normalized absolute POSIX paths. A rule on a folder applies to its descendants, and a rule may also be…
Seen at ElevenLabsThe code below uses OpenAI Chat Completions and is meant to run a short tool-calling exchange and return the model's final answer. Find the bugs. For each one, say what it…
Inspired by SierraNorthstar wants to move its agent-observability workflows to its existing ML platform and asks whether you support the move or recommend staying put. Before taking a position,…
Seen at LangChainAn enterprise customer wants its support agent to issue refunds instead of only recommending them. Before approving the capability or writing controls, identify roughly six…
Interview evidence from 2 companiesRecover exclusive function times from a nested stream of profiler events.
Seen at AnthropicWeekly active use of the assistant has fallen, and leadership wants a recovery plan next week. Before proposing one, start with about six checks that would show whether the…
Seen at GoogleBuild the scorecard you would use at the end of the beta to decide whether file uploads should expand to every paid account. Choose one north star and a compact set of…
Seen at xAIThe team must choose one workflow for the first field pilot. Set out the questions you would investigate with prospective users and the product team before making that choice.…
Seen at Google DeepMindDefine the scorecard you would bring to the six-week pilot review. Choose one north star and a compact set of supporting and countermetrics that show whether Study Mode…
Interview evidence from 2 companiesBefore proposing anything, set out around six discovery questions you would ask. For each, say what the answer would change about what you build. Do not pitch a solution yet.
Seen at ElevenLabsA label-quality dashboard ranks labeler 17 as the worst annotator in the current batch, and the team wants to exclude all of their labels from tomorrow's retrain. Start with…
Seen at OpenAI