ByteDance Interview Guide

ByteDance runs the most competitive-programming-flavoured loop of the companies here. The algorithmic bar is high and the problems skew harder than the big-five median -- expect genuine hard problems, tighter time limits, and follow-ups that go deeper into complexity rather than sideways into systems.

The loop also varies by region more than most: the process, the number of rounds and the emphasis differ between the China, Singapore and US pipelines. Ask your recruiter which one you are in.

Pitfalls

• Preparing for medium difficulty. The distribution genuinely skews harder here. If your hard-problem count is low, that is the gap.
• Being imprecise about complexity. Follow-ups drill into the bound rather than into system design. Know the difference between amortised and worst case, and which variable your bound is over.
• Assuming a standard loop. Round counts and emphasis differ substantially by region. Ask.
• Neglecting implementation speed. With harder problems in the same time box, fluency with heapq, bisect, itertools and the DP templates matters more than at companies with an easier distribution.

Interview follow-ups

They ask	What they're checking	The answer
"Can you do better than $O(n^2)$?"	Algorithmic depth	Name the specific bottleneck rather than saying yes. If the sort dominates, say so; if the transition is the problem, name the structure that removes it -- a monotonic deque, a segment tree, a prefix-sum trick.
"Prove that bound."	Precision under pressure	For an amortised bound, give the potential argument: each element is pushed once and popped once, so total inner-loop work is bounded by n. Nesting is not the bound.
"What is the complexity in terms of the constraint, not $n$?"	Whether you name the right variable	Binary search on the answer is $O(n \log R)$ in the value range; knapsack is $O(nW)$ in the capacity. Quoting either as a function of n alone is the slip they are probing.
"Which region's loop are you in?" (you asking)	Basic diligence	Ask the recruiter. The pipelines differ enough that it changes what to expect.

Recall card

• Hardest algorithmic distribution here. If your hard-problem count is low, that is the gap.
• Complexity follow-ups go deeper, not sideways. Amortised versus worst case, and which variable the bound is over.
• Be able to prove an amortised bound -- pushed once, popped once, total work bounded by n.
• Implementation fluency matters more -- harder problems in the same time box.
• The loop varies by region. Ask which pipeline you are in.

Self-check

quizCheck yourself

1. How does the ByteDance problem distribution differ from the big-five median?

   AIt is only system designBIt has no codingCIt is easier and more behaviouralDIt skews harder and more competitive-programming-flavoured
   Show answer

   D — It skews harder and more competitive-programming-flavoured — If your hard-problem count is low, that is the gap to close.

2. Binary search on the answer over a value range R with n elements costs O(n log R). What is the follow-up probing?

   AWhether you name the right variable: the bound is in the value range, not n aloneBWhether you know Big-Theta notationCWhether you can avoid binary searchDWhether you memorised it
   Show answer

   A — Whether you name the right variable: the bound is in the value range, not n alone — Knapsack is O(nW) in the capacity, for the same reason. Quoting these as functions of n alone is the slip.

3. How do you justify an amortised bound on a monotonic-stack loop?

   AThe nested loops multiply to O(n^2)BEach element is pushed once and popped once, so total inner work is at most nCThe stack is smallDAmortised means average over random inputs
   Show answer

   B — Each element is pushed once and popped once, so total inner work is at most n — Nesting is not the bound. Counting pushes and pops gives O(n) total.