Question 1

Context windows in 2025-26: how do you use 128k-1M tokens without drowning?

Accepted Answer

**Target Audience:** Engineers who have a 1M-token model available and need to decide when to use the window and when to keep retrieving.
**Difficulty:** Easy
**Verified with:** Node 22.22.0 and Python 3.12.3 (cost arithmetic, chunk reordering and a needle-test harness with canned answers; no real model is called). Model window sizes, prices and published guidance are cited inline with the date checked (October 2026); these change often, so check the provider docs for the model you use.

> **How

Question 2

What is a context window and what happens when you exceed it?

Accepted Answer

**Target Audience:** Beginners building chat features, RAG or agents who have hit (or will hit) "context too long" errors.
**Difficulty:** Easy
**Verified with:** Node 22.22.0 and Python 3.12.3 (a simulated provider with a 200-token window; no real model is called). Provider behaviour is cited inline from documentation and forum reports, checked October 2026.

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

Question 3

Embeddings in 2025-26: how do you choose model, dimensions, and distance?

Accepted Answer

**Target Audience:** Engineers setting up retrieval who must commit to an embedding model, a vector size and a metric, and want a defensible way to choose rather than copying a blog post.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (the truncation and quantization program below, on seeded synthetic vectors). Model facts, prices and limits are cited inline with the source and the date checked (October 2026); they change often, so treat them as examples of the mechanis

Question 4

How does similarity search work — cosine, dot product and Euclidean?

Accepted Answer

**Target Audience:** Engineers who have embeddings and now need to find the closest ones, and who want to know what the metric setting on their vector index actually does.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (both examples below: the three metrics worked by hand, and a small IVF index measured against exact search on seeded synthetic data). Database defaults and operator names are cited inline with the date checked (October 2026).

> **How to read this doc:**

Question 5

Reranking in RAG: when and how to add a cross-encoder?

Accepted Answer

**Target Audience:** Engineers who already understand what a reranker is and now have to decide whether to put one into a real RAG system, which one, how deep, and inside what latency budget.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (the depth sweep and timeout simulation below). The relevance scores are simulated with a seeded generator and the reranker latency comes from a stated cost model, so the numbers show the *method*, not a benchmark of any product. Vendo

Question 6

What is reranking and why add it after retrieval?

Accepted Answer

**Target Audience:** Engineers who have a working vector search or RAG prototype and keep hearing that "adding a reranker" is the cheapest quality win, and want to know what that actually means.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (the two-stage retrieval demo below). The reranker in the demo is a hand-written pair scorer standing in for a model, so it shows the *shape* of reranking, not the quality of a real cross-encoder. Product facts are cited inline with

Question 7

Evaluating RAG with Ragas: faithfulness, answer relevancy, context precision?

Accepted Answer

**Target Audience:** Engineers who run or are about to run Ragas on a RAG system and want to know what each number means, how it is computed, and what to fix when it moves.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (a hand-computed version of the four formulas below). Metric definitions were checked against the Ragas documentation pages for v0.2 to v0.4 via web search in October 2026; API details that only appeared in third-party guides are marked as such.

> **How

Question 8

Chunking in RAG: how do you pick size, overlap, and strategy?

Accepted Answer

**Target Audience:** Engineers who know what chunking is and now have to choose real parameters for a RAG index, and defend the choice with numbers in an interview.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (a sliding-window chunker swept over 12 size and overlap settings, scored by a TF-IDF retriever on 8 labelled questions; no model is called). Capacity arithmetic was run in Python 3.12.

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

Question 9

What is hybrid search and why combine BM25 with vector search?

Accepted Answer

**Target Audience:** Engineers whose embedding-only RAG search misses exact terms, and anyone who needs to explain lexical vs semantic retrieval and why teams combine them.
**Difficulty:** Medium
**Verified with:** Node 22.22.0 and Python 3.12.3 (BM25 written from scratch, a deterministic stub embedding, and RRF fusion on an eight-article knowledge base; no model or search engine is called). Engine facts were checked in each vendor's documentation in October 2026.

> **How to read this doc:** Ev

Question 10

How do you evaluate a RAG pipeline end to end?

Accepted Answer

**Target Audience:** Engineers responsible for a RAG system in production who need an evaluation that says *where* it fails, not only how often, and candidates facing a design-style evaluation question.
**Difficulty:** Hard
**Verified with:** Node 22.22.0 and Python 3.12.3 (the stage-by-stage harness below). Framework descriptions (TruLens RAG triad, Ragas) were checked against their docs via web search in October 2026.

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

AI ENGINEERINGDECODED.

Easy

Context windows in 2025-26: how do you use 128k-1M tokens without drowning?

What is a context window and what happens when you exceed it?

Medium

Embeddings in 2025-26: how do you choose model, dimensions, and distance?

How does similarity search work — cosine, dot product and Euclidean?

Reranking in RAG: when and how to add a cross-encoder?

What is reranking and why add it after retrieval?

Evaluating RAG with Ragas: faithfulness, answer relevancy, context precision?

Chunking in RAG: how do you pick size, overlap, and strategy?

What is hybrid search and why combine BM25 with vector search?

Hard

How do you evaluate a RAG pipeline end to end?

Why are BLEU and ROUGE poor metrics for LLM outputs, and what replaced them?

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

pgvector vs Pinecone vs Qdrant: how do you choose a vector DB in 2025?

How do agents maintain memory across steps and sessions?