swadhin / faang-playbook public
/ faang-playbook / 00-Hiring-Pipeline / 05-Onsite-Loop.md
swadhinbiswas feat: add Netflix L5 interview playbook · 2/2/2026
41c1cbb
121 lines · 4.2 KB
Raw Download

Onsite / Virtual Loop

The onsite (or virtual onsite) is usually 3–6 rounds, often combining:

  • 2× coding
  • 1× system design (mid/senior)
  • 1× behavioral
  • 0–2× role-specific (ML design, SQL, infra, domain)

This is where companies spend real money: multiple senior interviewers + a debrief process. The loop is designed to reduce false positives (hiring the wrong person) more than to avoid false negatives.

What companies want from the loop

They want multiple independent measurements of the same underlying capability, then a calibration step.

That’s why you’ll see similar themes across rounds (clarity, correctness, tradeoffs, execution).

In other words: you’re not trying to be “brilliant,” you’re trying to be consistently strong.

Before the loop (prep that actually matters)

1) Standardize your interview operating system

  • A repeatable coding process (clarify → plan → implement → test).
  • A repeatable system design process (requirements → API/data → components → scaling → reliability → tradeoffs).
  • A repeatable behavioral structure (STAR/SAO with metrics and ownership).

2) Build a “portfolio” of evidence

Have 6–10 stories ready:

  • 2 delivery wins
  • 2 reliability/incident stories
  • 1 conflict / disagreement
  • 1 leading without authority
  • 1 big technical tradeoff
  • 1 failure + what you learned

3) Match archetype expectations

  • Big tech: strongest on fundamentals + structured communication.
  • Startups: strongest on end-to-end ownership and pragmatism.
  • Enterprise: strongest on stakeholder management + maintainability.
  • Automotive/industrial: strongest on safety, validation mindset, constraints.

How to pace yourself

  • Early rounds: don’t “warm up.” Treat round 1 as the decider.
  • Between rounds: write 5 bullets of what went well/what to fix.
  • Food/water: manage energy like a long incident.

Add two more:

  • Reset between rounds: close tabs, breathe, sip water, quick posture reset.
  • Track time inside rounds: if you’re 10 minutes in and still unclear, force clarity.

Round-by-round guidance

Coding rounds

What gets you “strong hire” most often:

  • Clarifying questions early (inputs/outputs, constraints)
  • Clear plan before coding
  • Clean implementation
  • Systematic tests (including edge cases)
  • Able to explain complexity and tradeoffs

System design (mid/senior)

What kills candidates:

  • Building before requirements
  • No data model, no APIs, no capacity planning
  • No failure handling (timeouts, retries, backpressure)
  • No tradeoff discussion

Behavioral

What kills candidates:

  • Vague stories, no ownership
  • No metrics, no conflict, no learning
  • Blaming others

If the loop is at Amazon (or similar rubric-heavy companies)

Some companies are very explicit about “loop interviews” and structured hiring. Amazon publishes a “How We Hire” overview that highlights loop interviews and preparation resources.

Common “loop-killers”

  • Great solution, poor communication (interviewer can’t follow).
  • Incorrect code + weak testing.
  • System design without requirements and constraints.
  • Behavioral answers without ownership, conflict, or measurable outcomes.

Also common:

  • Over-optimizing a subcomponent while ignoring the main requirement.
  • Arguing with the interviewer instead of collaborating.
  • Not correcting yourself when you notice a bug.

Cross-links

Grounding sources