Kakao data scientist SQL and coding interview 2026

How many interview rounds does Kakao's Data Scientist hiring process have?

The process consists of four distinct rounds executed over a 21‑day window, and the final decision is made by a cross‑functional hiring committee (HC).

In Q2 2026, after the third interview, the hiring manager raised a timing objection because the candidate had taken 12 days to submit the coding assignment. The HC forced a 48‑hour deadline for the final take‑home, illustrating that the schedule is not flexible. The judgment is clear: treat the four‑round timeline as immutable.

The first round is a 45‑minute screening with a senior data scientist focusing on product intuition. The second round is a live SQL deep‑dive lasting 60 minutes, where the interviewers probe data modeling, window functions, and performance tuning. The third round is a 90‑minute algorithmic coding session on a whiteboard or shared editor, emphasizing time‑complexity analysis. The fourth round is a take‑home case study, delivered within 48 hours, that combines SQL extraction with a Python‑based predictive model.

The hiring committee evaluates each round independently but aggregates signals into a single “fit” score. The committee’s final vote is binary: “hire” if the composite exceeds the threshold, “reject” otherwise. The decision is not swayed by a single strong performance; the candidate must demonstrate consistency across all rounds.

Insight: The process follows a “signal stacking” framework—each round contributes a weighted signal, and the HC applies a majority‑vote rule. Candidates who over‑prepare for one round risk neglecting others. The correct strategy is balanced preparation across all four stages.

What SQL problems actually appear in Kakao's data scientist coding interview?

Kakao tests practical data‑engineering scenarios rather than textbook query rewrites; expect multi‑table joins, hierarchical aggregations, and performance‑aware indexing questions.

During a recent debrief, the panel highlighted a candidate who answered a textbook “top‑10 customers” query flawlessly but failed to explain why a GROUP BY on a high‑cardinality column caused a full table scan. The HC noted that the candidate’s “SQL correctness” was acceptable, but “operational awareness” was absent. The judgment: mastery of SQL syntax is not enough; you must demonstrate query‑optimization thinking.

A typical problem asks candidates to retrieve the last purchase date per user, then compute a 30‑day rolling churn metric. The solution requires a self‑join or window function, followed by a WHERE clause that filters on DATE_DIFF. The interviewers also ask candidates to rewrite the query using a materialized view to improve latency, testing knowledge of Kakao’s data warehouse (Hive on Presto).

Another frequent scenario involves time‑travel queries on a partitioned log table, where candidates must write a query that returns events occurring before a given timestamp while avoiding partition‑pruning pitfalls. The interviewers look for a WHERE event_time < ? clause combined with explicit partition filters, ensuring the engine can prune early.

Insight: Kakao’s “SQL realism” framework expects candidates to treat queries as production code, not isolated exercises. The not‑“memorized syntax” but “performance‑aware design” contrast recurs throughout the interview.

Which coding language and algorithm focus should I prioritize for Kakao?

Python is the default language for Kakao’s data‑science coding interview, and the algorithmic emphasis is on graph traversals, dynamic programming, and probabilistic modeling rather than classic sorting problems.

In a Q3 debrief, a senior data scientist complained that a candidate spent the entire 60‑minute coding slot implementing quick‑sort, while the interview prompt explicitly required a shortest‑path algorithm on a weighted graph. The HC recorded a “misaligned focus” flag, resulting in a reject despite flawless syntax. The judgment: align your language and algorithm choice with the problem domain, not with personal comfort.

The interview prompt often describes a recommendation system: given a bipartite user‑item interaction matrix, compute the top‑K items for a user using collaborative filtering. Candidates must implement a matrix factorization routine or a simple cosine‑similarity calculation, then evaluate it with a metric like MAP@K. The interviewers assess both the algorithmic correctness and the ability to vectorize operations using NumPy or Pandas.

A second common problem asks candidates to simulate a Monte‑Carlo process that estimates the probability of a user churn event based on historical features. The solution requires random sampling, convergence checking, and a runtime‑complexity explanation. The interviewers probe for understanding of bias‑variance trade‑offs and the impact of sample size on confidence intervals.

Insight: Kakao follows a “domain‑driven algorithm” model: the chosen algorithm should solve a real data‑science problem, not a generic CS puzzle. The not‑“generic algorithm” but “business‑aligned solution” distinction is decisive.

How does the hiring committee evaluate “product sense” for a data scientist at Kakao?

Product sense is measured by the candidate’s ability to translate data insights into actionable product recommendations, and it carries a 30 % weight in the HC’s final score.

During a debrief after the take‑home case study, the HC debated a candidate who delivered a statistically perfect churn model but failed to suggest a feature rollout plan. The senior product manager on the committee argued that “the model is a tool, not a decision.” The HC voted to downgrade the candidate’s overall fit, demonstrating that product sense outweighs pure technical excellence. The judgment: you must close the loop from analysis to product impact.

The product‑sense interview is the first 45‑minute screen, where interviewers present a product metric (e.g., daily active users) and ask the candidate to hypothesize a root cause for a dip. Candidates must ask clarifying questions, propose data‑driven experiments, and estimate the impact of potential A/B tests. The interviewers score the candidate on hypothesis formulation, experiment design, and communication clarity.

A recurring mistake is to answer with a single‑metric analysis (e.g., “user retention dropped because sessions decreased”). The correct approach is multi‑dimensional: examine cohort retention, funnel conversion, and external events. The not‑“single‑metric focus” but “holistic product narrative” approach distinguishes high‑performers.

Insight: The “product‑impact lens” framework forces candidates to think beyond data extraction; they must demonstrate how insights drive product decisions. The HC’s 30 % weighting makes this a decisive factor.

What compensation can I realistically negotiate after a Kakao data scientist offer?

Base salary typically ranges from $130,000 to $170,000, with an additional 0.04 %–0.08 % equity grant and a signing bonus between $15,000 and $30,000; negotiation success hinges on the strength of the HC’s “fit” signal.

In a 2026 negotiation debrief, a candidate leveraged a strong take‑home performance to secure a $165,000 base, a 0.07 % equity tranche, and a $28,000 signing bonus. The HC noted that the candidate’s “fit” score was in the top‑10 % of the cohort, permitting a higher leverage margin. The judgment: compensation is proportional to the HC’s confidence, not to market benchmarks alone.

When discussing equity, be prepared to reference Kakao’s recent valuation: a $12 billion market cap at the time of the offer. A 0.05 % grant translates to roughly $6,000 in immediate value, but the HC expects candidates to understand vesting schedules (four‑year vest with a one‑year cliff). The not‑“equity as a perk” but “long‑term wealth component” mindset is critical during negotiations.

Signing bonuses are typically allocated for candidates who must relocate to Seoul within 30 days. If you can demonstrate a higher cost of living or a competing offer, you can push the bonus toward the upper bound. The HC will not increase the base salary beyond the range without a compelling signal of market scarcity.

Insight: Kakao’s “signal‑linked compensation” model ties financial terms to the HC’s composite score. Candidates who excel in product sense and SQL performance can extract the maximum package, while those who underperform in any round will see a compressed offer.

Preparation Checklist

Review Kakao’s public data‑pipeline architecture (Hive + Presto) and practice writing partition‑pruned queries.
Solve three graph‑traversal problems in Python, focusing on Dijkstra and BFS implementations with heapq.
Build a end‑to‑end churn prediction notebook: data cleaning, feature engineering, model training, and MAP@K evaluation.
Memorize the syntax for window functions (ROW_NUMBER() OVER (PARTITION BY … ORDER BY …)) and practice optimizing them with indexes.
Simulate a 48‑hour take‑home case study, then iterate based on feedback; incorporate product recommendation language.
Work through a structured preparation system (the PM Interview Playbook covers Kakao‑specific SQL performance and product‑sense frameworks with real debrief examples).
Prepare a concise product‑impact story: pick a past project, quantify the business outcome, and rehearse a 2‑minute pitch.

Mistakes to Avoid

BAD: Submitting a perfectly correct SQL query but ignoring execution plans. GOOD: Explain why the query may trigger a full scan and propose an index or materialized view.

BAD: Spending the entire coding interview on a classic sorting algorithm that does not address the problem domain. GOOD: Align the algorithm with the business case, e.g., use Dijkstra for shortest‑path recommendations.

BAD: Offering a data‑driven insight without a concrete product experiment. GOOD: Pair each insight with an A/B test design, expected lift, and measurement plan.

📖 Related: Kakao data scientist interview questions 2026

FAQ

What is the typical timeline from application to offer for Kakao’s data scientist role?

The end‑to‑end process averages 21 days, with four interview rounds spaced roughly every 4–5 days, followed by a 48‑hour take‑home and a final HC decision within two business days.

Do I need to know Korean to succeed in Kakao’s data scientist interviews?

Technical interviews are conducted in English; however, the product‑sense screen may reference Korean user metrics, so basic familiarity with Korean terminology helps but is not mandatory.

Can I negotiate equity after receiving an offer, or is the base salary the only lever?

Equity is negotiable, especially if your HC “fit” score is in the top‑10 %. Present a clear case linking your technical impact to long‑term company growth, and you can increase the equity grant by up to 0.03 % of the company.

Ready to build a real interview prep system?

Get the full PM Interview Prep System →

The book is also available on Amazon Kindle.

TL;DR

Review Kakao’s public data‑pipeline architecture (Hive + Presto) and practice writing partition‑pruned queries.