Amazon hires more interns globally than almost any other tech company, but most candidates miss that SDE Intern and DA (Data Analyst) Intern run on two completely different OA + VO pipelines. Same amazon.jobs portal, same resume - the SDE pipeline is Work Simulation + Coding + Workstyle + Debug, while the DA pipeline swaps in SQL Lab + Statistics MCQ + Case Study. Studying for SDE while interviewing for DA (or vice versa) almost always produces a bad result.
This article compares the two tracks module by module, with a representative real question and solution per track. By the end you should be able to answer two things:
- Which track does my current resume actually fit?
- What should I drill in each phase, and where does each track usually filter people out?
Same Entry, Different Funnel
| Dimension | SDE Intern | DA Intern |
|---|---|---|
| Application portal | amazon.jobs / campus events | amazon.jobs / campus events |
| Teams | AWS / Alexa / Retail / Ads | Finance / FBA / Marketing Analytics |
| OA platform | HackerRank + proprietary | HackerRank + proprietary |
| OA modules | Work Simulation + 2 Coding + Workstyle + Debug | SQL Lab + Stats MCQ + Case Study + Workstyle |
| VO rounds | 3-4 rounds: algorithm + LP x2 + System Design Lite | 3-4 rounds: SQL + Case Study + LP x2 |
| Decisive skill | Algorithms + LP narrative | SQL speed + business intuition + LP narrative |
One-line summary: SDE evaluates code fundamentals plus LP, DA evaluates SQL speed plus business reasoning plus LP. LP is the only common denominator.
SDE Intern OA: Four Modules
Module 1: Work Simulation
Simulates a workday packed with emails, meetings, and decisions. The format is multiple choice and probes prioritization under pressure, customer obsession, and data-driven thinking. Common traps:
- When "customer experience" and "cost" both appear as options, customer experience comes first.
- Coworker asks for help vs your own deadline -> answer with communication cadence, not direct refusal.
Module 2: Coding (Representative: Sum of Skills team assignment)
Given an array of N employees with skill values, split them into groups of size K such that the variance of group skill totals is minimized.
Approach: sort and pair greedily - pair the strongest with the weakest, second strongest with second weakest, and so on.
def team_assignment(skills, K):
skills.sort()
n = len(skills)
teams = []
i, j = 0, n - 1
while i < j:
team = []
while len(team) < K and i <= j:
team.append(skills[j]); j -= 1
if len(team) < K and i <= j:
team.append(skills[i]); i += 1
teams.append(team)
return teams
Time complexity: O(N log N), dominated by the sort. Follow-up: if the prompt becomes "make every group's skill sum equal to a target," it collapses into multi-knapsack - mind the data size.
Module 3: Workstyle Survey
No right or wrong answers, but internal consistency is scored. Tips:
- Keep "Strongly Agree" / "Strongly Disagree" usage under about 30 percent.
- "I prefer working alone" is a trap - Amazon prefers collaboration.
Module 4: Debug
About 15 minutes. You receive a small program with 5-7 bugs and must fix them. Common bug families:
- off-by-one
- using
iswhere==is required - mutating a dict while iterating
DA Intern OA: Four Modules
Module 1: SQL Lab (Representative: cohort retention)
Given
orders(user_id, order_date, amount), compute monthly new-user retention over the next 3 months.
WITH first_order AS (
SELECT user_id, MIN(order_date) AS first_dt
FROM orders
GROUP BY user_id
),
cohort AS (
SELECT
DATE_TRUNC('month', first_dt) AS cohort_month,
user_id
FROM first_order
)
SELECT
c.cohort_month,
COUNT(DISTINCT c.user_id) AS new_users,
COUNT(DISTINCT CASE
WHEN o.order_date BETWEEN c.cohort_month + INTERVAL '1 month'
AND c.cohort_month + INTERVAL '3 month'
THEN o.user_id END) AS retained_users
FROM cohort c
LEFT JOIN orders o USING (user_id)
GROUP BY c.cohort_month
ORDER BY c.cohort_month;
Key points:
DATE_TRUNCdiffers slightly between Amazon Redshift and open source Postgres - watch the dialect.- Define retention as
COUNT(DISTINCT ... CASE WHEN ...)in a single pass.
Module 2: Statistics MCQ
10-15 multiple choice items covering:
- Hypothesis testing: two-sample t-test, chi-square
- A/B Testing: power analysis, MDE
- Probability: conditional probability, Bayes
- Regression: multicollinearity, heteroskedasticity
Trap: options often equate "statistically significant" with "business significant." Wrong.
Module 3: Case Study
Business prompts. Common archetypes:
- "FBA Returns are up 8 percent. Analyze in a week."
- "Prime Day GMV missed plan. Diagnose."
Answer skeleton (SDE-W: Segment / Drill / Experiment / Wrap):
- Segment: by category, region, customer segment
- Drill: find anomalous subset
- Experiment: design A/B to verify
- Wrap: one-sentence stakeholder conclusion
Module 4: Workstyle - same as SDE.
VO Loop Comparison
| Round | SDE Intern | DA Intern |
|---|---|---|
| 1 | LP x2 + Coding Easy | LP x2 + SQL x1 |
| 2 | Coding Medium + LP x1 | SQL Hard + Case Study |
| 3 | System Design Lite + LP x1 | Case Study + LP x2 |
| 4 (some teams) | Bar Raiser: full LP | Bar Raiser: full LP |
Bar Raiser is identical for both tracks - LP-driven, 5-6 stories, each story drillable into specific behaviors.
Preparation Roadmap Comparison
| Dimension | SDE Intern | DA Intern |
|---|---|---|
| Primary platform | LeetCode (Amazon tag) | LeetCode SQL + StrataScratch |
| Practice volume | 200+ problems | SQL 100 + stats 50 |
| Stories needed | 6-8 LP stories | 6-8 LP stories |
| Mock interviews | Karat / Pramp | DataLemur / alumni mocks |
Three-Week Cadence (SDE)
- Week 1: LeetCode Amazon tag, top 50 mediums
- Week 2: 10 hard coding problems plus the P-A-S system design template
- Week 3: LP story polishing plus 2 Karat mocks
Three-Week Cadence (DA)
- Week 1: StrataScratch Amazon SQL top 50
- Week 2: An A/B Testing textbook plus 3 case study sets
- Week 3: LP stories plus 2 DataLemur mocks
FAQ
Q1: I applied to both SDE and DA Intern - which OA do I get? Recruiters route based on the "primary tone" of your resume. If both signals are strong, the JD you clicked through wins - that track sends the OA.
Q2: Is DA Intern SQL easier than SDE Coding? No. DA SQL Hard often layers window functions inside CTEs, and 80 minutes for 4 problems is tighter than LeetCode's 60-minutes-for-2 cadence.
Q3: Does Workstyle Survey filter candidates? Not on its own, but combined with weak OA performance it can become a tiebreaker. Stay consistent and avoid extreme picks.
Q4: Is the Bar Raiser at OA or Onsite? Onsite. There is no Bar Raiser at OA. The Bar Raiser is a cross-team senior with veto power in the final loop.
Q5: If OA is mediocre, can I still get to VO? SDE leans on coding; VO ordering tracks OA scores. DA evaluates SQL plus case study together. Referrals plus resume highlights help, but treat OA as the primary gate either way.
Preparing for Amazon Intern OA / VO?
If you have an OA link but the Work Simulation prioritization is uncertain, you cannot finish 4 SQL problems in 80 minutes, or you want a real person doing VO proxy / VO assist on Bar Raiser day, we can talk through a complete OA proxy / VO assist / VO proxy plan.
Contact
Need real interview questions and a custom prep plan? Add WeChat Coding0201 now to get questions.
Email: [email protected] Telegram: @OAVOProxy