One-line summary: TikTok 2026 SDE NG = OA (CodeSignal 4 questions) + Phone Screen (1 question) + 4 Onsite rounds (2 Coding + 1 SD + 1 BQ/HM). Algorithms lean LeetCode Hot 100, mostly Medium difficulty — efficiency is everything.
TikTok (ByteDance's U.S. arm) has been one of the most active NG hirers over the past two years, with bases in Mountain View / San Jose / Seattle. Interview cadence is fast, offers come quickly, and comp is competitive. This article documents the latest 2026 reports.
1. TikTok NG Pipeline
| Stage | Format | Length | Content |
|---|---|---|---|
| Resume Screen | — | 1 week | Initial filter |
| OA | CodeSignal | 70 min | 4 algorithm questions |
| Phone Screen | Lark video | 60 min | 1 Coding + light BQ |
| Onsite Round 1 | Lark video | 60 min | Coding (DS + algo) |
| Onsite Round 2 | Lark video | 60 min | Coding (DP / graph) |
| Onsite Round 3 | Lark video | 60 min | System Design (NG-lite) |
| Onsite Round 4 | Lark video | 45 min | Hiring Manager + BQ |
| Offer Call | Phone | 30 min | Comp negotiation |
Pace: OA → offer typically takes 3-5 weeks — about 2-3× faster than Meta / Google.
2. OA: CodeSignal 4 Questions
Q1: Equation Division (LC 399 variant)
Statement: Given a/b = k, compute c/d.
Idea: model variables as graph nodes, weighted edges, BFS / DFS for path product.
from collections import defaultdict, deque
def calcEquation(equations, values, queries):
g = defaultdict(dict)
for (a, b), v in zip(equations, values):
g[a][b] = v
g[b][a] = 1 / v
def bfs(s, t):
if s not in g or t not in g:
return -1.0
q = deque([(s, 1.0)])
seen = {s}
while q:
x, val = q.popleft()
if x == t: return val
for y, w in g[x].items():
if y not in seen:
seen.add(y)
q.append((y, val * w))
return -1.0
return [bfs(s, t) for s, t in queries]
Q2: Rating with Memory Prefix
Statement: Streamed user ratings. Each step output the rolling average over the past K ratings.
Idea: deque + running sum (avoid recomputation).
Q3: Top K Frequent Hashtags
Statement: Maintain top K hashtags by frequency with add(tag) and query(k).
Idea: dual HashMap + bucket sort (lazy delete).
Q4: Random Set Design
Statement: O(1) insert / remove / getRandom (LC 380 variant).
Idea: HashMap + array (swap-with-last on remove).
3. Phone Screen: 1 LeetCode Medium
| Pattern | LeetCode analog | Frequency |
|---|---|---|
| Two pointers + sliding window | LC 3 / 76 | ⭐⭐⭐⭐⭐ |
| Tree traversal | LC 102 / 124 | ⭐⭐⭐⭐ |
| Graph traversal | LC 200 / 207 | ⭐⭐⭐ |
| String matching | LC 28 / 49 | ⭐⭐⭐ |
Common follow-ups:
- What if the array is too large for memory?
- What if it's streaming data?
- Can you optimize the time complexity further?
4. Onsite 4 Rounds
Round 1 — Coding (Array + Two Pointers)
Sample: Find all triplets with arr[i] + arr[j] + arr[k] == target (LC 15 variant).
Follow-up: What if the array has 1B elements? → External Sort + two pointers.
Round 2 — Coding (DP / Graph)
Sample: Cheapest flights within K stops (LC 787) — minimum-cost path with at most K edges.
Idea: Bellman-Ford variant or Dijkstra with state (node, steps).
Round 3 — System Design (NG-lite)
Common questions:
- Design TikTok feed recommendation
- Design Like / Follow service
- Design short-video upload + transcoding pipeline
NG expectations:
- Draw high-level architecture (API Gateway, Service, DB, Cache)
- Identify bottlenecks (hot keys, read-heavy)
- Provide basic numbers (QPS, storage, cache hit rate)
Tip: NG SD doesn't need Paxos / Raft, but always clarify scope first.
Round 4 — Hiring Manager + BQ
| Question | Expected emphasis |
|---|---|
| Tell me about yourself | 30-second elevator pitch |
| Why TikTok? | ByteDance algo platform / recsys |
| Most proud project | Quantified gain (QPS, users, revenue) |
| A failure | Highlight reflection + growth |
| A conflict | STAR structure |
TikTok values:
- ByteStyle: Always Day 1
- Be Candid and Clear
- Be Open and Humble
- Champion Diversity and Inclusion
5. TikTok NG Salary Range (2026)
| City | Base | Stock (4 yr ADR) | Sign-on | Total / Yr |
|---|---|---|---|---|
| Mountain View | $155-175K | $80-120K | $20K | $200-225K |
| San Jose | $150-170K | $80-120K | $20K | $195-220K |
| Seattle | $145-165K | $70-110K | $15K | $185-210K |
Note: TikTok-US stock is typically ByteDance ADR-style (cash-equivalent); actual value depends on vesting and IPO timing.
6. FAQ — TikTok NG Common Questions
Q1: Are interviews in Chinese or English?
Both work. Interviewers usually ask "prefer Chinese or English." Pick whatever you express better in.
Q2: Lark or CoderPad?
OA on CodeSignal, Onsite on Lark Meeting + embedded coding pad (CoderPad-like but lighter).
Q3: When does H1B sponsorship happen?
After interviews, visa status confirmed. New H1B lottery in March-April. If unselected, OPT/STEM OPT extends work eligibility.
Q4: Does TikTok mind LeetCode original problems?
No, but interviewers tweak I/O and add follow-ups.
Q5: How long until I can re-apply?
3-6 months. Make sure your skills have visibly improved.
Q6: Is TikTok NG good for non-CS majors?
Yes. TikTok accepts ECE / Math / Physics, but you should have 1-2 software projects on your resume.
7. 4-Week Sprint Plan
| Week | Focus | Daily output |
|---|---|---|
| 1 | LC Hot 100 + TikTok company tag | 5-7 |
| 2 | DP + graph + design | 5 + 1 SD |
| 3 | System Design + BQ stories | 3 SD + 8 STAR |
| 4 | Full mock + real questions | 1 end-to-end mock |
8. External Resources
🚀 Need TikTok OA / VO Coaching?
If you're preparing for TikTok, ByteDance, Lemon8, or CapCut NG / intern interviews, we can help with CodeSignal mocks, onsite coding lists, System Design, and BQ story polishing.
👉 Add WeChat: Coding0201 — get real questions and a 1-on-1 prep plan
Contact
- WeChat: Coding0201
- Email: [email protected]
- Telegram: @OAVOProxy