Katkat
Back

Build and present a minimal vehicle log playback app from S3 data in two hours

Applied IntuitionSoftware Engineer · Onsite
MediumSoftware EngineerOnsiteSoftware Engineering Fundamentals

Last updated

Given two hours, an AI coding assistant and an S3 bucket of unfamiliar vehicle log data, build a minimal 3D log playback app and then present it in 30 minutes. The exercise tests data exploration, tight scoping, verifying AI-generated code, and explaining trade-offs and future work clearly.

This round follows a system design round in which you designed a real-time 3D playback viewer for recorded vehicle logs. You now get read access to an S3 bucket of recorded log data and two hours to build a minimal working version of that app. The company provides an AI coding assistant for the build. After the two hours, you have 30 minutes to present the project.

Constraints and Clarifications

  • Time box: two hours of building, then a 30-minute presentation.
  • The format and size of the files in the bucket are not described up front; finding out is part of the task.
  • The AI coding assistant is available throughout the build.

Clarifying Questions

  • Must the app run in a browser, or is a local desktop or notebook viewer acceptable?
  • May you download files from the bucket to your machine, and are some of them small enough to inspect quickly?
  • What counts as done: one drive playing end to end, or seek and several data channels as well?
  • Will the presentation include a live demo, a code walkthrough, or both?

Part 1 — Build a minimal playback app in two hours

Plan and carry out the build. Explain how you would find out what the logs contain, what minimal version you would aim for first, how you would split the work between yourself and the AI assistant, and in what order you would add features as time allows.

What This Part Should Cover

  • Inspecting a small sample of the data before writing viewer code
  • A minimal end-to-end slice delivered early, then features added in priority order
  • Directing the assistant in small, checkable steps and verifying its output
  • Carrying the design round's on-demand loading and nearby caching into the build, or a stated reason for deferring it

Part 2 — Present the project in 30 minutes

Present what you built: what it does, how it is structured, which trade-offs you made under the time limit, and what you would develop next.

What This Part Should Cover

  • A clear walkthrough of the architecture and data flow, backed by a demo
  • Explicit trade-offs made under the time limit and their consequences
  • A prioritized plan for future development that ties back to the design round
  • An honest account of what you verified yourself and what the assistant generated

Hints

Hint 1

Look before you build: The bucket's contents are unknown. Decide the cheapest way to learn the schema and the timing of the data before any viewer code exists.

Hint 2

Cut scope by the clock: Decide what your demo must show at the one-hour mark if everything after that goes wrong.

Hint 3

Prepare the story early: Before time runs out, decide what written artifact would let you walk through the project calmly instead of reading code live.

Hint 4

Name what you skipped: For every shortcut you took, be ready to say what it costs and what you would build instead.

Follow-up questions

  • Your prototype keeps a whole drive in memory. What breaks first when a drive is an hour long, and what would you change?
  • Which part of the assistant-generated code do you trust least, and how would you test it?
  • How would you turn this prototype into a tool several engineers use every day?