Company interview guide
Block (Square) Machine Learning Interview Guide (2026)
Live pairing on practical Python and SQL, then a long onsite of hands-on modeling, case studies and ML Q&A.
Block (Square)'s machine learning interview loop has 8 reported stages, from recruiter call to behavioral and hiring-manager wrap-up; typical length: 2 weeks to 3 months.
Roles covered: Data Scientist and ML Engineer. The question bank holds 9 questions reported from Block (Square) interviews, 1 of them free to solve in your browser.
Formerly Square. Square, Inc. renamed itself Block in December 2021; Square is now one of its brands. Reports tagged Square are shown here.
- Interview process length
- 2 weeks to 3 months
- Stages
- 8
- Reports linked to bank questions
- 9
- Bank report dates
- Feb 2014–May 2025
- Questions in bank
- 9 (1 free)
- Last reviewed
- October 2026
How the Block (Square) interview loop works
A recruiter call, a technical screen, a hiring-manager call, then a virtual onsite of about five or six interviews (four to six hours, sometimes split over two days). Data scientist screens are usually 30 minutes of SQL plus about 45 minutes of Python; ML engineer screens are one pairing or algorithm coding round, and one Cash App MLE in 2024 got a take-home instead. Onsites mix a live data exploration or modeling round on a provided dataset, product, A/B testing or business cases, ML design or ML question-and-answer rounds, coding for engineers, and behavioral time with the hiring manager. A 2025 ML engineer offer described four onsite rounds (two coding, one design, one behavioral) with a lead and a shadow interviewer in each. Python is expected even when a posting mentions R. Several candidates found a round described as modeling or exploration closer to a take-home squeezed into an hour, so practise working fast on unfamiliar data.
The Block (Square) interview process, stage by stage
01Recruiter call
About 30 min by phone or video
A conversation about your background, why Block and the team, and how the rest of the loop runs. Some recruiters slip in one or two quick technical checks, such as how to get the unique values of a dataframe column or what a SQL window function does. Recruiters often send detailed preparation notes, including which coding topics to practise.
What it tests
- A clear summary of your most relevant project
- Why Block, Square or Cash App specifically
- Quick SQL or pandas basics in some calls
- Logistics: level, location, timeline, language for the coding rounds
How to prepare
- Confirm the coding language; several candidates who planned to use R were told the rounds are in Python.
- Ask for the round list and the preparation guide; candidates say the recruiter's notes match what follows.
- Have a one-line answer ready for a SQL window function and for deduplicating a column.
Recruiter or HR call opens the process in 18 data and ML candidate reports from 2013 to 2025, four of them since 2024; two mention quick SQL or pandas questions.
02Technical screen
Data science: about 30 min SQL plus about 45 min Python, live with video on; ML engineering: one pairing or algorithm coding round of about 45-60 min
For data scientists the screen is usually two short live rounds: basic SQL (group by, distinct counts, joins, a window function) and Python that leans on strings, dictionaries, tuples and nested loops rather than libraries; some Cash App screens use pandas group-bys instead. ML engineer screens are a single pairing exercise, such as a class that grows as requirements change or a sliding-window average over a stream, or a standard medium algorithm question such as the kth largest element. One 2024 Cash App MLE was given a take-home in place of a live screen.
What it tests
- Plain Python on strings and dictionaries without pandas
- SQL aggregation, joins and window functions written quickly and correctly
- Pandas group-by and filtering on a small table
- Adapting working code when the interviewer changes the rules
- Testing and explaining your own code while pairing
How to prepare
- Drill dictionary and string manipulation in pure Python; these screens are easy but timed and scored closely.
- Clarify the expected output before you write SQL; candidates say prompts were brief on purpose.
- In pairing rounds, get a working version first, then improve it and say how you would test it.
- Keep the camera on; one recruiter called a missing video feed a red flag.
Free reported questions for this stage
- Moving Average Over a Data StreamCoding & AlgorithmsMediumSolve
Screen described in 20 data scientist and 8 ML engineer reports, seven since 2024; DS screens are SQL plus Python in most. The company's engineering blog describes its live pairing format.
03Hiring manager call
About 30-60 min, usually after the screen and before the onsite; some candidates had it first
A conversation with the manager about your past work and the role. For ML engineers it can be a technical deep dive into a model you built: architecture, evaluation, metrics and tooling. Some managers add ML technical questions here.
What it tests
- Depth on one past project, from data to evaluation
- Model choice and metrics you used, and why
- Fit with the team's product and problems
- Motivation for this team
How to prepare
- Pick one project and know its metrics, baselines and trade-offs cold.
- Read up on the product the team owns; one candidate's recruiter advised exactly this.
Hiring-manager call before the onsite in 10 data and ML reports, five since 2024; two ML engineer reports describe a technical project deep dive.
04Onsite: live data exploration and modeling
1-2 x about 50-60 min on your own machine with a provided dataset, screen shared; SQL may be written on a described schema without an IDE
The most distinctive round. You get an unfamiliar dataset and must explore it, clean it and build and evaluate a model within roughly an hour, then discuss metrics. A separate round may focus on data manipulation or exploration questions, or on SQL pairing. Candidates repeatedly said these feel like take-homes compressed into an hour, and one who prepared a fixed modeling template found it easy.
What it tests
- Fast exploration and cleaning of an unfamiliar table
- An end-to-end model: preprocessing, split, fit, validate
- Choosing and explaining evaluation metrics
- Data manipulation and SQL under time pressure
How to prepare
- Prepare a reusable notebook skeleton: load, inspect, clean, baseline, model, evaluate, summarize.
- Ship a simple baseline in the first 20 minutes, then improve it.
- Ask questions about the data early; some interviewers want you to move on rather than explore at length.
Live modeling or data rounds in 12 data and ML reports, four since 2024, consistently on a provided dataset in about an hour; two describe SQL pairing or schema-only SQL.
05Onsite: product, A/B testing and business cases
1-2 x 30-60 min discussions, some with a product or business partner
Data scientist loops nearly always include case work: designing and reading an A/B test, a churn or customer analysis, product metrics, or how you would run a project with a stakeholder. ML modeler loops have described a modeling case alongside an experimentation case. One 2025 Cash App data scientist called the case harder than at most companies.
What it tests
- A/B test design, metrics and interpretation
- Churn, segmentation and product-metric reasoning
- Framing a modeling approach for a payments or risk problem
- Working with stakeholders and prioritizing requests
How to prepare
- Practise A/B test design end to end, including sample size and pitfalls.
- Use Cash App and Square products as your examples: sellers, payments, peer transfers, fraud.
- State the decision the analysis supports before you choose metrics.
Case, experimentation or business-partner rounds in 14 data and ML reports, five since 2024; A/B testing is named in five and churn or product metrics in three.
06Onsite: ML design and ML Q&A
1-3 x about 60 min whiteboard or discussion rounds, mainly for ML engineer and ML modeler roles
ML engineer loops include design and question-and-answer rounds: designing an ML system for a banking problem from input data to production, explaining why one model fits a problem better than another, and ML fundamentals and statistics. One 2025 MLE got a general system design prompt (a hotel reservation system) rather than an ML one. Expect follow-ups that go beyond the literal question.
What it tests
- End-to-end ML system design for a fintech problem
- Model selection and the reasoning behind it
- ML basics and probability or statistics questions
- Serving, scaling and API design for a model
How to prepare
- Prepare a fraud or risk-model design: labels, features, imbalance, thresholds, monitoring.
- Cover what the interviewer did not ask too; one candidate was marked down for missing unprompted topics.
- Revise core probability (distributions, covariance, transforms) for the statistics module.
ML design, ML Q&A or statistics rounds in 7 ML engineer and ML modeler reports, five since 2024; prompts vary by team and one was a general system design.
07Onsite: coding (ML engineers)
1-2 x about 60 min pairing or algorithm rounds, each with a lead and a shadow interviewer
ML engineer onsites include one or two coding rounds of medium difficulty, either as pair programming or as a standard algorithm question. A 2025 offer described two such rounds in a four-round loop. Data scientist onsites usually skip this round.
What it tests
- Medium-difficulty algorithms: arrays, heaps, dynamic programming, linked lists
- Clean, runnable code written together with the interviewer
- Testing and explaining trade-offs as you go
How to prepare
- Use the tools and language you know best and run your code often.
- Talk through trade-offs out loud; both interviewers take notes.
Onsite coding in 4 ML engineer reports (2021-2025), as pairing or medium algorithm rounds; the company's engineering blog describes pairing in general.
08Behavioral and hiring-manager wrap-up
1-3 x 30-60 min at the end of the onsite, often including the hiring manager again or a senior leader
Conversations on how you work: conflict, prioritizing stakeholder requests and leadership, plus a review of past experience. Some loops end with a second hiring-manager talk or a senior leader.
What it tests
- Handling conflict with colleagues or stakeholders
- Prioritizing competing requests
- Ownership and impact on past projects
How to prepare
- Prepare stories about conflict and prioritization with concrete outcomes.
- Keep energy up; this comes at the end of a four to six hour day.
Behavioral, leadership or final hiring-manager rounds in 8 data and ML reports, two since 2024; conflict and stakeholder prioritization are the named topics.
Questions and themes reported most often
- SQL basics: group by, distinct counts, joins and a window function (8 reports, technical screen)
- Build and evaluate a model on an unfamiliar dataset in about an hour (6 reports, onsite: live data exploration and modeling)
- Pure-Python string, dictionary and tuple manipulation with loops (6 reports, technical screen)
- Design and interpret an A/B test (5 reports, onsite: product, a/b testing and business cases)
- Deep dive on a past model: architecture, evaluation, metrics (5 reports, hiring manager call)
- Pandas group-by and filtering questions (3 reports, technical screen)
- Classification metrics and model choice (3 reports, onsite: ml design and ml q&a)
- Churn or customer analysis case (2 reports, onsite: product, a/b testing and business cases)
What Block (Square) asks about
- Python & algorithm coding25%
- Live modeling & ML Q&A25%
- SQL & pandas20%
- Product, A/B testing & business cases20%
- Behavioral & hiring manager10%
Block (Square) interview prep checklist
- Solved a timed set of SQL questions with joins and a window function
- Did pure-Python string and dictionary exercises in 45 minutes
- Cleaned and modeled an unfamiliar dataset end to end in under an hour
- Can design and interpret an A/B test for a payments product
- Prepared a fraud or credit-risk ML design for ML engineer loops
- Know one past project's metrics and trade-offs in depth
- Confirmed the language, round list and onsite length with the recruiter
Tips
- Write all coding answers in Python unless told otherwise; R users were caught out more than once.
- Drill fast, plain-Python data wrangling with dictionaries and strings, not just algorithm puzzles.
- Build a modeling template you can run from memory on any table in under an hour.
- Practise SQL with window functions and joins until you can write them without running them.
- Prepare an A/B test walkthrough and a churn or fraud case grounded in Square and Cash App products.
- For ML engineer roles, add medium algorithm practice and one fintech ML system design.
- Ask clarifying questions early; several candidates found prompts deliberately short.
- Plan for a four to six hour onsite and ask whether it can be split across days.
See the rest of the Block (Square) loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Block (Square) loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Block (Square) interview FAQ
How many interview rounds does Block (Square) have for ML roles?
Candidates report 8 stages: Recruiter call, Technical screen, Hiring manager call, Onsite: live data exploration and modeling, Onsite: product, A/B testing and business cases, Onsite: ML design and ML Q&A, Onsite: coding (ML engineers) and Behavioral and hiring-manager wrap-up. A recruiter call, a technical screen, a hiring-manager call, then a virtual onsite of about five or six interviews (four to six hours, sometimes split over two days). Data scientist screens are usually 30 minutes of SQL plus about 45 minutes of Python; ML engineer screens are one pairing or algorithm coding round, and one Cash App MLE in 2024 got a take-home instead. Onsites mix a live data exploration or modeling round on a provided dataset, product, A/B testing or business cases, ML design or ML question-and-answer rounds, coding for engineers, and behavioral time with the hiring manager. A 2025 ML engineer offer described four onsite rounds (two coding, one design, one behavioral) with a lead and a shadow interviewer in each. Python is expected even when a posting mentions R. Several candidates found a round described as modeling or exploration closer to a take-home squeezed into an hour, so practise working fast on unfamiliar data.
How long does the Block (Square) ML interview process take?
Reported timelines run 2 weeks to 3 months.
What does Block (Square) ask in ML interviews?
By reported question mix: Python & algorithm coding 25%; Live modeling & ML Q&A 25%; SQL & pandas 20%; Product, A/B testing & business cases 20%; Behavioral & hiring manager 10%. The most repeated themes: SQL basics: group by, distinct counts, joins and a window function; Build and evaluate a model on an unfamiliar dataset in about an hour; Pure-Python string, dictionary and tuple manipulation with loops.
How should I prepare for the Block (Square) ML interview?
Write all coding answers in Python unless told otherwise; R users were caught out more than once. Drill fast, plain-Python data wrangling with dictionaries and strings, not just algorithm puzzles. Build a modeling template you can run from memory on any table in under an hour.
Are there free Block (Square) interview questions?
Yes. LeetML has 9 questions reported from Block (Square) interviews, and 1 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- StripePractical multi-part coding, a timed dataset-to-model round, an ML bug squash and payments-flavoured ML design.
- PayPalSQL and Python screens, medium-difficulty algorithm coding, ML fundamentals and fraud-flavoured cases, with loops that vary by team.
- CoinbaseCognitive and culture tests up front, hard SQL, live modeling on messy data, and a work-trial presentation for data scientists.
- RobinhoodSQL on trading data, probability and churn modeling for data scientists; coding, ML design and a project deep dive for ML engineers.
- AffirmPractical, job-shaped ML interviews: EDA, feature engineering, metrics for credit and fraud models, plus medium-level coding.
- Capital OneA timed four-question coding assessment, then a four-round Power Day of ML technical, a business case, behavioral and, on recent ML engineer loops, ML system design.