xAI interview reports in this collection cover an ML-engineering phone screen, a software-engineering interview, frontend prompts, and a take-home assessment. They do not describe one shared interview loop. This guide keeps each role signal separate and leaves short or incomplete problem descriptions at the level the candidate report supports.
How this page was researched
This page summarizes xAI entries in Parlel's LeetCode Discuss research collection, prepared on October 31, 2026. The collection does not preserve a publication year for every post. The links below point to the candidate threads; where a source notes only a topic, UI requirement, or assessment format, the guide does not invent omitted constraints.
Roles represented in the reports
| Role or stage in the source | Reported material | What is missing |
|---|---|---|
| ML engineer phone interview | A sized LRU cache task | The collection does not preserve a complete transcript |
| Software engineer interview | Nested-list iterators, LLM token parsing, and typeahead design | Full constraints are not retained for every task |
| Frontend interview | Debounced typeahead, a growing-sequence array task, and a prompt-history interface | One array task has only a brief label in the notes |
| US take-home | Four hours and a choice of six prompts | The prompt statements were not shared in the thread |
| CodeSignal assessment | Assessment-process discussion | No question content was captured |
These examples can help you practice relevant skills, but a role title alone does not prove that the same exercise will appear in another interview.
ML engineering phone screen: sized LRU cache
An ML-engineer phone interview report names a Sized LRU Cache. Unlike a cache limited by item count, this version gives each entry a size; when a new item would exceed the total capacity, least-recently-used entries are evicted until it fits. The task is described in the Grok/xAI ML engineer phone-interview report on LeetCode Discuss.
For practice, clarify how an entry larger than the entire capacity should behave, whether updating a key changes its size and recency, and whether a read or write counts as use. Those are useful specification questions, not extra interviewer prompts attributed to the report.
Software engineering interview: coding and design
A separate xAI interview thread lists several tasks:
- Create iterators over a list and nested lists. The collection preserves this short task description but not the full interface. (LeetCode Discuss: xAI interview)
- Parse tokens for an LLM. The report notes the task theme without enough detail to reconstruct token rules or expected output. (same xAI report)
- Design a typeahead system. (same xAI report)
Keep the iterator and parsing prompts at this level unless the original report supplies more detail. Replacing missing constraints with a familiar standard problem would make the summary less reliable.
Frontend interview prompts
The xAI frontend report describes three different exercises:
- Build a typeahead with debounce and highlight the matching text in filtered results. (LeetCode Discuss: xAI frontend interview)
- Find the length of a subsequent growing sequence in an array. The research notes retain this brief wording only, so this guide does not assume it means a particular longest-increasing-subsequence variant. (same frontend report)
- Build a two-pane prompt interface: prompt history on the left and context plus input on the right; fetch a response for a new prompt or replay a history prompt with its earlier response as context. (same frontend report)
The interface task is especially useful to scope before coding. Ask how history is selected, whether requests can overlap, how loading and errors appear, and what data replay sends. These are implementation questions for practice, not additional requirements reported by the candidate.
Take-home and online assessment notes
One US take-home thread describes a four-hour exercise with a choice among six prompts but does not include their statements. The xAI CodeSignal thread also discusses an assessment without sharing the questions. Neither source supports a question list, so this page records only the process notes: xAI take-home discussion and xAI CodeSignal OA discussion.
Preparation based on the reports
- Implement an LRU cache with both item-count and total-size capacity, explaining the invariant and eviction behavior.
- Practice nested iteration and make the iterator's state explicit before writing
nextandhasNextbehavior. - For typeahead, cover debouncing, request cancellation or stale responses, keyboard interaction, and safe text highlighting.
- For the prompt interface, separate history state from the active request and make replay behavior testable.
- Ask the recruiter whether the role includes live coding, system design, a take-home, or an assessment; the reports show that the formats vary.
Frequently asked questions
Are these questions specific to xAI?
They are tasks described in candidate reports that identify xAI interviews. The page does not claim they are official, repeated, or exclusive to xAI.
What are the xAI take-home prompts?
The linked thread says candidates could choose from six prompts during a four-hour take-home, but it does not publish their contents. This guide leaves them unnamed.
Does xAI use one loop across software and ML roles?
The available reports cover different roles and stages. They do not establish a universal process.
Compare the OpenAI software engineer interview reports, review system design interview questions, or return to the company interview experience hub.
Looking for a referral at xAI?
Create a profile on Parlel to unlock hiring manager emails for roles you’re targeting.