Amazon · Big Data Engineering · Fifty questions, defensible decisions, visible proof

Interview Q&A

Practice 50 Amazon-like e-commerce big-data questions across scale, Kafka, CDC, streaming, the lakehouse, modeling, finance, serving, ML, governance, privacy, recovery, and cost.

50-question interview workspace

Defend the complete Amazon big-data design

Eight rounds cover sources, scale, Kafka, CDC, streaming, the lakehouse, analytical modeling, finance, serving, ML, governance, and recovery. Select a question to see the position, trade-off, and proof an interviewer can challenge.

QUESTION 01 / 50

Sources + scale

Prime Day is approaching. Which Amazon commerce systems produce the evidence your data platform must capture?

Amazon scenario to walk through

A two-day global campaign spans Storefront, Search, Ads, Cart, Orders, Payments, Inventory, Fulfillment, Seller Central, Customer Service, fraud systems, carriers, and payment processors. Identify the authority and capture mode for each rather than calling everything clickstream.

Amazon-specific interview answer

Orders, payments, inventory, fulfillment, catalog, pricing, promotions, sellers, customer identity, search, recommendations, advertising, web/mobile behavior, support, fraud, and external carriers or processors all produce independent evidence. I separate authoritative business events and CDC from high-volume behavioral telemetry and partner feeds because their guarantees, privacy, ordering, and failure modes differ.

Trade-off to defend

A single universal event bus looks simple but couples unrelated traffic and ownership; domain topics add operational overhead but preserve failure isolation and meaning.

Evidence or validation

For every producer, record its owner, transaction boundary, event grain, key, capture mode, volume, privacy class, and recovery source.

Interview note: numerical values are visible planning assumptions for an Amazon-like platform, not claims about private production infrastructure.

Five-minute whiteboard

Build the Amazon answer in six connected decisions

Move from the transaction boundary to scale, sources, Kafka and streaming, trusted products, governance, and recovery. Hover, focus, or tap a step to see what to draw and explain.

STEP 01Frame the boundary0:00–0:40

Draw two boxes: the transactional commerce plane and the big-data plane. Put checkout, payment authorization, inventory reservation, and fulfillment commands in the first; put events, Kafka, streaming, Iceberg, analytics, and ML in the second.

Explain the decision

State the invariant first: analytics may become stale, but it never authorizes payment, grants inventory, changes an order, or owns fulfillment. Operational services commit first and hand off evidence asynchronously through outbox/CDC or governed telemetry.

The interviewer should know what is in scope and what must remain online before any technology choice appears.

Preparing chapter

Loading your data design lesson

The curriculum shell is ready while the requested chapter is being prepared. You can wait here or return to the design library.

Open design library

Continue learning

NetflixStreaming analytics platform

UberMarketplace data platform

YouTubeVideo analytics platform