Question 1

How do you get reliable structured JSON output from an LLM?

Accepted Answer

**Target Audience:** Engineers whose LLM feature has to hand JSON to ordinary code (a parser, a database, an API) and who want it to stop breaking.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (the constrained-decoding simulation below). Provider API shapes were checked against the vendor docs and announcements cited inline, in October 2026.

> **How to read this doc:** Every idea is explained in plain language first, then tagged with `📌 Interview term:` so you also

Question 2

When do you use code-based graders versus model-based graders?

Accepted Answer

**Target Audience:** Engineers writing their first eval suite who need to decide, check by check, whether to write a function or ask a model, and candidates who should be able to defend that choice.
**Difficulty:** Medium
**Verified with:** Node 22 and Python 3.12 (five code graders and a stub judge over ten outputs, and a judge-versus-human validation with corrected pass rates). Judge verdicts are canned stand-ins for model calls; token prices in the cost line are assumptions. Vendor tooling is

Question 3

What is LLM-as-judge and what are its known biases?

Accepted Answer

**Target Audience:** Engineers about to let a model grade their model, and candidates who need to name the biases and the fixes precisely.
**Difficulty:** Medium
**Verified with:** Node 22 and Python 3.12 (a pairwise judge with a simulated position bias, and a scoring judge with a simulated verbosity bias). The judges are deterministic stubs with a bias built in on purpose, so the mechanism and the fix are visible; the size of the bias is an assumption, not a measurement of any model. Published

Question 4

How do you handle PII and sensitive data when calling LLM APIs?

Accepted Answer

**Target Audience:** Engineers putting a hosted LLM behind a product that touches customer data (support, CRM, HR, health, finance) and who need a design legal and security can sign off.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (the redaction pipeline below). Provider retention terms were checked against the Anthropic docs and OpenAI's data-controls guide as summarised in search results, in October 2026. This is engineering guidance, not legal advice.

> **How to

Question 5

What are guardrails and how do you implement them around a model?

Accepted Answer

**Target Audience:** Engineers adding an LLM feature to a product who need to explain, and build, the checks that run around the model call.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (both examples below, identical output). Product facts (Llama Guard 4, Prompt Guard 2, NeMo Guardrails rail types, Guardrails AI on-fail actions, the OpenAI moderation endpoint, Amazon Bedrock ApplyGuardrail) were checked by web search in October 2026 and are cited where they appear.

Question 6

What is semantic caching and when is it safe to use?

Accepted Answer

**Target Audience:** Engineers with repetitive LLM traffic (support bots, FAQ assistants, internal help desks) who have been told a semantic cache will halve the bill, and want to know when that is true and when it serves wrong answers.
**Difficulty:** Hard
**Verified with:** Node 22.22.0 and Python 3.12.3 (a semantic cache with safety rails, and a threshold sweep on labelled pairs). The embedding in both examples is a **toy stand-in** (words mapped to concepts) so the similarity numbers are exp

Question 7

MCP authoring: how do you build a Model Context Protocol server in 2025-26?

Accepted Answer

**Target Audience:** Engineers asked to expose a product or internal system to AI hosts through MCP, and candidates who need to talk through the build end to end.
**Difficulty:** Hard
**Verified with:** Node 22.22.0 and Python 3.12.3 (both examples, identical output). SDK facts were read in October 2026 from the published packages: `@modelcontextprotocol/server` 2.3.1, `@modelcontextprotocol/core` 2.3.1, `@modelcontextprotocol/node` 2.1.1, `@modelcontextprotocol/sdk` 1.32.1 and the Python `mcp`

Question 8

What are common RAG failure modes and how do you debug them?

Accepted Answer

**Target Audience:** Engineers who run or are about to run a RAG feature in production and need a systematic way to turn "the bot got it wrong" into a specific, fixable cause.
**Difficulty:** Hard
**Verified with:** Node 22.22.0 and Python 3.12.3 (the chunk-boundary demo and the triage harness below). The request logs in the triage are hand-written examples of each failure type, not data from a real system. Research findings are cited inline.

> **How to read this doc:** Every idea is explained

Question 9

How do you validate and repair structured LLM output at runtime?

Accepted Answer

**Target Audience:** Engineers putting LLM output into systems of record (databases, billing, tickets) who need a runtime layer that catches bad output and fixes what can be fixed.
**Difficulty:** Hard
**Verified with:** Node 22.22.0 and Python 3.12.3 (two programs: the validate-repair-fallback pipeline and the truncation demo). Library and provider behaviour is cited inline with the source, checked October 2026.

> **How to read this doc:** Every idea is explained in plain language first, then

AI ENGINEERINGDECODED.

Medium