Datadog’s AI coding interview: official rules and candidate reports

Published 4 November 2026 · 2 min read

Datadog is one of the few companies with a public page on AI in its interviews. For some roles there’s an AI coding interview where AI is expected, and you’re told about it in advance. Everywhere else, AI is off. This guide combines that official page with candidate reports from May to October 2026.

The official rules

Datadog’s careers page, updated 13 May 2026, sets out where AI is and isn’t allowed (Interviewing at Datadog: AI Guidelines):

• AI coding interviews: AI is allowed and expected for certain roles. “You will be explicitly told in advance if you will be in an AI coding interview.”
• Other live coding and assessments: no AI, “unless explicitly stated otherwise”.
• Non-technical interviews: no AI; using it “may result in your application being disqualified”.
• Disclosure: “Misuse of AI tools, or failure to disclose their use, may lead to disqualification.”

What the AI round assesses, in Datadog’s words: “how effectively you leverage AI to solve problems, reason through solutions, and make engineering decisions”. Be ready to explain how and why you used the AI.

Where it sits in the loop

• A July poster on Blind said the onsite had been updated to include an LLM-oriented system design round and an AI-assisted coding round, “rather than traditional Leetcode” (Blind).
• A senior candidate in August had a virtual onsite of 4 interviews: AI coding, system design, behavioural and values (Blind).

Two formats: feature or code review

A candidate in May was offered a choice before the round: code a feature, or do an AI-assisted code review (Blind). The most detailed report covers the feature-and-bugs version:

• A hosted dev environment with an existing Go codebase and Claude Code or Codex.
• The task: implement a feature and find some bugs.
• The agent does the typing. The candidate was expected to have the agent do all the coding and debugging. They worked entirely in the Claude Code command line, plus a second terminal for their own commands, with no IDE.
• What mattered: familiarity with concurrency and distributed systems. They didn’t know Go, and with an agent that wasn’t a problem. This was a staff loop that ended in an offer (Blind).

We haven’t found a detailed first-hand account of the code review version yet. If you’ve had it, add your report.

How to prepare

1. Confirm the format with your recruiter: AI coding or not, feature or code review, and which agent.
2. Practise driving a coding agent from the terminal, in Claude Code or Codex, on a codebase you don’t know.
3. Brush up on concurrency and distributed-systems basics: race conditions, retries, timeouts, batching.
4. For code review, practise reviewing AI-written changes: what breaks under load, what’s untested, what you’d ask the author.
5. Explain as you go. Datadog says it evaluates your reasoning and decisions, not just the result.

Our free practice projects give you an unfamiliar codebase with a bug, a feature and a speed-up, and our guide to the code review interview covers that format in general. All Datadog reports with sources: Datadog reports page.

Sources

• Datadog careers: Interviewing at Datadog: AI Guidelines (updated 13 May 2026)
• Blind thread, May 2026: choice of feature or code review
• Blind thread, posted 26 July: onsite updated with AI coding and LLM system design
• Blind thread, August 2026: senior virtual onsite
• Blind thread, October 2026: staff loop, Go codebase with Claude Code or Codex