Back to Behavioral
Behavioral
hard
mid

Walk me through your project responsibilities and illustrate the overall architecture with a diagram.

This is the classic 'walk me through a project' opener. The interviewer wants a structured narrative: business context, your role and ownership, system architecture (with a diagram), the key trade-offs you made, the hard problems you hit, and what you would do differently. Treat it as a 5-minute pitch with optional deep-dives. The diagram is the anchor — practice drawing one cleanly.

7 min read·~5 min to think through

What the interviewer is actually asking

This prompt looks open-ended but it has a clear shape:

  1. What was the project? (1 sentence business context)
  2. What was your role? (scope and ownership)
  3. What did the system look like? (the diagram)
  4. What were the hard parts? (technical depth)
  5. What did you learn / would change? (reflection)

They want to see structured thinking, clear ownership, and the ability to talk at multiple levels of abstraction.

Structure your answer (5 minutes)

1. Set the stage — 30 seconds

"At [company], I led the frontend for [product]. It served ~X users, with [N] engineers. My team owned [scope]; I was [tech lead / principal / IC] specifically responsible for [your slice]."

This anchors scale, your role, and what's yours to talk about.

2. Business problem — 30 seconds

"The core problem was [X]. We needed to [Y] under [constraint Z]."

Constraints are the seed of every interesting design decision. Surface them up front.

3. Architecture diagram — 90 seconds

Draw on the whiteboard or sketch verbally:

ts
┌─────────────┐    ┌──────────────┐    ┌─────────────┐
│  Browser    │───▶│  Edge / CDN  │───▶│   API GW
│  Next.js    │    │  (Vercel)    │    │   (Node)    │
└─────────────┘    └──────────────┘    └──────┬──────┘

                              ┌───────────────┼───────────────┐
                              ▼               ▼               ▼
                        ┌──────────┐   ┌──────────┐    ┌──────────┐
                        │ Postgres │   │  Redis   │    │  S3
                        └──────────┘   └──────────┘    └──────────┘

Narrate it: "Request hits Vercel edge for static + SSR. Dynamic routes call API gateway. Postgres is the primary store, Redis for sessions and cache, S3 for assets."

4. The hard parts — 90 seconds

Pick 2-3 trade-offs or problems and explain how you decided:

  • "Tab data was growing 10x year-over-year, so we moved from in-memory to cursor pagination + virtualization, which dropped TTI from 3s to 800ms."
  • "Auth was originally in the gateway, but mobile and web needed different flows, so we moved it to a dedicated identity service."
  • "Caching at the CDN was tricky because user content was personalized — solved with stale-while-revalidate + edge functions for the personalization layer."

Numbers + decisions land best.

5. What you'd change — 30 seconds

"In retrospect, we should have [pulled X out earlier / not built Y / picked Z]. The lesson was [...]."

This shows reflection without being self-deprecating.

Diagram principles

  • Boxes for systems, arrows for direction of data flow.
  • Label arrows with protocol (HTTP, WebSocket, SSE, gRPC).
  • Color or group by tier (client / edge / server / data).
  • Include scale numbers if you know them.
  • Practice this before the interview — drawing a clean diagram on a whiteboard is a skill.

What to AVOID

  • Buzzword soup ("we used micro-services, CQRS, event sourcing, DDD..."). Pick what mattered.
  • "We" overload — own your part. The interviewer needs to know what YOU did.
  • Going off on tangents — keep deep-dives for follow-up questions.
  • Missing the business context — engineers love jumping to tech.
  • No numbers — "fast", "scalable", "lots of users" tells them nothing.

Anticipate follow-ups

The interviewer will pull one thread to depth. Prepare:

  • "Why Postgres over [X]?"
  • "How did you handle [auth / caching / state]?"
  • "Walk me through a tricky bug you debugged on this system."
  • "How did the architecture evolve over time?"
  • "What would breaking changes have looked like?"

For each major component, know one war story.

Adapting by seniority

  • Mid: emphasize ownership of a specific feature, what you built, decisions you made.
  • Senior: emphasize cross-team coordination, trade-offs, mentoring.
  • Staff/Principal: emphasize org-level impact, strategy, what you de-prioritized and why.

Mental model

A project walkthrough is a 5-minute pitch with deep-dives on demand. Structure: context → role → diagram → 2-3 hard problems with decisions and numbers → reflection. The diagram is the anchor for everything else. Practice the narrative; don't wing it.

Follow-up questions

  • Walk me through one specific architectural trade-off and your reasoning.
  • What was the hardest bug you debugged on this system?
  • How did the architecture evolve as you scaled?
  • What would you do differently if you started over?

Common mistakes

  • Skipping the business context and jumping into tech.
  • Buzzword-laden architecture descriptions with no concrete decisions.
  • 'We' overload — interviewer can't tell what YOU owned.
  • No diagram — purely verbal description loses the audience fast.
  • Going deep on every component instead of letting interviewer pull threads.

Performance considerations

  • Practice the narrative out loud, with a timer. 5 minutes feels short. Have a clean diagram you can draw in under 2 minutes. Memorize the 2-3 most interesting trade-offs.

Edge cases

  • Interviewer interrupts mid-explanation — pivot, don't reset.
  • You worked on a small slice of a huge system — frame your slice clearly.
  • Project was a failure — own it, explain what you learned.
  • Confidential project — describe shape and trade-offs without specifics.

Real-world examples

  • Staff-level interviews almost always open with this question.
  • FAANG L5+ rounds expect a 10-minute deep dive with diagram.
  • Startup interviews often want fewer slides, more raw decisions.

Senior engineer discussion

Seniors lead with constraints and trade-offs, not technology choices. They have a small set of well-rehearsed war stories that demonstrate scale, ownership, and judgment. They can move between altitudes — from a 10,000ft architecture view down to a specific debugging episode — without losing the thread, and they invite the interviewer to pull on whichever string is most interesting.

Related questions