OA format by company¶
For anyone who just got an OA invite and wants the format before opening it. Sixty companies, as of Oct 2026, from official pages and recent candidate reports. Formats vary by role, region, team and season.
How to use this page¶
- Find your company below. Read the Platform and Time columns first.
- Open that platform's section on OA platforms for scoring rules and free practice.
- Compare with your invite. The invite wins. If they disagree, the format changed.
- Read the last 3 months of reports for your company: the company's LeetCode Discuss topic (swap the company name in the URL, sort by newest) and its company page linked in the table.
- Ask your recruiter anything the invite leaves out: language limits, allowed docs, whether AI is allowed.
How to read the table:
- Official means the company's own careers page says it. Everything else is reported by candidates, with the month where it matters.
- Varies by role means sources disagree. The table gives the most recent sourced claim and flags the older one.
- Company names link to the company guide, which lists the sources and the most asked questions.
By platform¶
| Platform | Companies (as of Oct 2026) |
|---|---|
| HackerRank | Amazon, Microsoft, Goldman Sachs, JPMorgan, Citadel, Two Sigma, D. E. Shaw, Stripe, IBM, Salesforce, Cisco, Expedia, Adobe, Oracle, ServiceNow, Nutanix, Snowflake, DoorDash, Palantir, Rippling, Walmart, OpenAI, Agoda, Uber (2026 reports), Visa (from Feb 2026), Nvidia (when used), Hudson River Trading (or Codility) |
| CodeSignal | Capital One, Visa, Pinterest, TikTok, Meta (some roles), Netflix (intern reports), Airbnb, Anthropic, Coinbase, eBay, Robinhood, Instacart, Databricks (when used), Roblox (coding part), PhonePe (lateral) |
| Codility | Palo Alto Networks, Microsoft (earlier cycles), Hudson River Trading (or HackerRank), Google (some off-campus) |
| Other | Google Hiring Assessment (Google's own), SHL/AMCAT (Morgan Stanley India), Glider.ai (Intuit India), HackerEarth and Unstop (Flipkart), Karat (PayPal, Walmart US), Roblox's own games |
| Usually no OA | Bloomberg, Jane Street, Waymo, Tesla, Snap, Spotify, Cloudflare (most tracks), Anduril, Apple (many US teams), DoorDash (many US SWE I), Nvidia (many US teams) |
FAANG¶
| Company | Platform | Questions | Time | Extra sections | Notes |
|---|---|---|---|---|---|
| Amazon | HackerRank, webcam on | Official: 2 coding problems. Reported since Apr to May 2026: 1 DSA problem plus 1 AI-assisted repo task (fix bugs in a Node/MERN, Spring Boot or Django app) | Official: coding about 70 min. Reported 2026: a coding challenge of about 100 min (40 to 45 DSA, 55 to 60 repo) | Workstyles (about 15 min, built on the Leadership Principles), feedback survey (about 5 min). Full-time only: Work Simulation (official average 60 min; the page also says "You are provided 120 minutes") | Official page grades "the clarity, maintainability, and efficiency of your code", and its FAQ says "A webcam is required at the start of the code debugging test". Intern page lists no Work Simulation, but 2026 intern reports include one (about 45 min) plus a Work Style survey (about 6 min): varies by role. Print screen ends the session. Avoid copy-paste (Amazon OA prep) |
| Google Hiring Assessment for most applicants (official). Coding OA on HackerRank, Codility or Google's platform for some | GHA: agree or disagree statements (Aced: 50 to 75 items). Coding: 1 to 2 problems | Coding 60 to 90 min (reported) | GHA measures "workstyle skills" (official) | Referral and recruiter-sourced candidates often skip the coding OA. India campus OAs: 2 hard problems with hidden tests. After a no, Google asks you to "wait at least a year" for the same type of role (Google how we hire) | |
| Meta | CodeSignal, proctored (camera, mic, screen) | New grad: 1 progressive problem in 4 stages, or 4 separate problems. Intern: 4 problems of rising difficulty | New grad 90 min. Intern about 70 min | Some candidates: a Likert-scale behavioral survey | Not everyone gets an OA. Meta's official first step for most SWE candidates is a 45-minute technical screen (Meta tech screen). Reference docs allowed, no AI. Most candidates do not finish all 4 stages |
| Apple | Varies by team: CoderPad or HackerRank-style | 2 coding, sometimes 15 to 20 MCQs on the team's stack | 60 to 70 min | Stack MCQs (JavaScript/React, Java) | Many US teams skip the OA and start with a recruiter or hiring manager call. Some IS&T roles fix the language to Java |
| Netflix | CodeSignal (2025 to 2026 student reports). Official first step for students: a "take-home assessment" | Intern (Oct 2025 reports): 4 coding questions, then a separate AI-assisted General Coding Assessment | 70 min for the coding part | AI-assisted GCA, scored separately | Netflix does not publish the format. ML and AI intern versions are more ML-flavored. Treat the assistant as a pair: plan, check its code, keep moving |
| Microsoft | HackerRank in most 2025 to 2026 reports. Codility in earlier cycles | Usually 2: one easier, one medium or hard. A Feb 2025 Codility OA (SDE II) had 3 in 60 min | 60 to 110 min (India campus intern: 75 min) | none | India campus OAs report tab-switch and paste detection. Passing every test does not guarantee a call. Official rule: no outside assistance "unless explicitly permitted" (Microsoft how we hire) |
Read: The Definitive Guide for Meta & Amazon Engineering Interviews: Jugal's Feb 2026 walkthrough of Amazon's three OA sections (coding, Work Simulation, Work Style). His tips for the two non-coding sections are on strategy.
Big tech¶
| Company | Platform | Questions | Time | Extra sections | Notes |
|---|---|---|---|---|---|
| Nvidia | HackerRank (official how-we-hire page names it for coding exercises). Campus tests in India | When used (Aced): a 75-min HackerRank test with at least 2 DSA questions plus MCQs. India campus: aptitude and C MCQs plus 2 coding | 60 to 75 min | Aptitude, C, pointers. One 2024 India test had negative marking (29 questions, 60 min) | Many US SWE candidates skip a standalone OA. Official: using "unapproved outside tools, such as ChatGPT" leads to disqualification (Nvidia how we hire) |
| Uber | HackerRank (2026 reports). CodeSignal historically | 2 to 4 | 65 to 90 min | none | Jan 2026: 90 min, 2 questions (medium plus advanced). A SWE I report (Jan 2026): a 65-minute round with MST, DSU and next-greater-element problems. US new grad reports on CodeSignal used the 4-question GCA shape. Partial solutions have advanced (3 of 4 solved) |
| Not named in reports | India intern and new grad: 2 to 3 DSA, one hard | about 90 min | none | US and experienced candidates usually start with a 60-minute CoderPad phone screen. Aug 2026 India OA needed multi-source Dijkstra | |
| Salesforce | HackerRank | AMTS: 3, the last one hard. MTS and SMTS: 2 medium | AMTS 90 to 100 min. MTS 75 min | none | A May 2026 C++ OA: 2 questions worth 50 and 75 marks, 80% to qualify |
| Adobe | HackerRank. Unstop for the India Hackathon | US intern and new grad (Nov 2024 report): 3 LeetCode mediums. India campus: 2 coding plus CS MCQs | Not stated | OS, DBMS, OOP MCQs (India) | FY27 intern candidates asked about a bug-fix and code-review item. Replies say mostly DSA plus small debugging |
| Oracle | HackerRank | India campus: 2 coding plus MCQ sections | Experienced: 60 min | Aptitude, reading, probability, OS, DBMS, networks, OOP, SQL, REST | Little US early-career data |
| Intuit | Glider.ai (India campus). Off-campus platform not named | Campus intern: 4 DSA. SE1 off-campus 2026: 1 DSA, 1 SQL, 1 Bash | Campus 90 min, proctored | SQL and Bash are scored | Candidates who skipped SQL or Bash asked if they could still pass. Do all three |
| PayPal | HackerRank. Karat for some US roles | Intern: 1 to 2 DSA plus CS MCQs. Early career: 2 to 3 DSA plus Spring Boot/Java MCQs, SQL or a class-design task | Intern 45 to 90 min | MCQs, SQL, machine coding | Mixed sections: time management decides the result |
| Visa | CodeSignal (GCA style, still reported Aug 2026) and HackerRank tiered tests (from Feb 2026) | CodeSignal: 4. HackerRank Expert: 3. New grad Sep 2026: 3 | CodeSignal 70 min (90 in one drive). HackerRank 90 min | none | 567/600 was rejected (Dec 2025). 600/600 did not guarantee an interview. Check your invite for the platform |
| IBM | HackerRank, proctored | US: 2 (easy plus medium). Some 2026 roles: 1 DSA plus 1 REST API task. Some report 3 | 45 to 90 min, 60 most common. Link valid 7 days | Some roles: recorded video and English tests. India: a second test with a framework bug-fix (Node.js, Spring Boot or Django) | IBM says AI "should not be used: To complete assessments, technical tests, or coding challenges" (IBM application process) |
| Cisco | HackerRank | US: 2 to 4 easy or medium. India campus (Code with Cisco): 42 questions including 2 long coding problems | about 90 min. US window about 2 weeks | India: MCQs and written CS questions | US OA is emailed right after you apply. Resumes are reviewed after the OA. Cisco lists "Generating the solution to a coding problem" in real time as poor AI use (Cisco GenAI practices) |
| Qualcomm | Proctored MCQ test (India campus). Many US loops have no OA | India campus Aug 2026: 60 MCQs in 3 sections | 90 min | C/C++ output prediction, bit manipulation, pointers, OS, architecture. +1 and -0.25 marking | Qualcomm's interview policy bans AI tools, bots and LLMs in assessments unless Qualcomm asks (Qualcomm interview policy) |
| Tesla | No standard OA | When present: 1 to 3 practical problems (for example a JSONPath evaluator) | Varies | none | Treat the first live technical round as the OA |
| Walmart Global Tech | HackerRank (2026 off-campus). HackerEarth or HirePro (India campus). Karat (many US first rounds) | 2026 HackerRank: 3 sections: 1 coding, 1 front-end task (React or Angular), 1 back-end task (Node, Spring or Django) | Not stated | Front-end and back-end sections allow an AI assist | Back-end applicants still get the front-end section (Mar 2026). India campus 2024: 22 MCQs plus 2 coding in 75 min |
| eBay | CodeSignal, GCA or ICA | GCA: 4 (easy, easy, medium, medium). ICA: 1 project in 4 levels | GCA 70 min. ICA 90 min | none | Recruiters may accept an existing CodeSignal score. Deep copies can fail ICA level 4 performance tests |
| Expedia | HackerRank | 3: usually 1 easy plus 2 medium, sometimes medium-hard | 90 min (one report: 105) | Emerging Talent: strength-based and skills assessments (official) | Solving all 3 does not guarantee a call (Aug 2025 report: "Not Selected"). Statements are long; read the constraints (n up to 10^6). Up to 5 applications per 45 days (Expedia FAQ) |
| ServiceNow | HackerRank | 2 to 3 (medium, medium, hard) | 60 to 90 min | QA intern OA added Selenium, SQL, Jenkins MCQs and a prompt-writing task | Official: complete tests without AI unless told otherwise |
| Palo Alto Networks | Codility plus HireVue (US new grad 2025). Varies by country | US Master's-level role (Aug 2026): 3 (1 basic, 2 intermediate). Israel: 3 easy to medium, on site | 75 min (US, Aug 2026) | HireVue video (US new grad 2025). India: aptitude and MCQs | You are told in advance if an assessment applies (official FAQ) |
| Atlassian | HackerRank (India 2026 grads). An Atlassian-branded IDE (2027-grad India OA, Aug 2026) | 2 to 4 DSA. India campus around Codeforces 1600 | about 90 min | UK graduate track: psychometric tests first (Sep 2026) | The OA is mandatory for early careers (Atlassian early careers guide). Experienced roles report a Karat screen instead |
| Spotify | Not standard | When used: LeetCode easy to hard, sometimes a logic puzzle | n/a | Data roles: SQL | Most candidates go straight from the recruiter call to a live screen |
| CodeSignal GCA | Standard GCA: 4 | 70 min | none | Official: a passing score triggers the recruiter call. An ML intern advanced with 550. Confirm the question count in your invite | |
| Snap | Not consistently reported | n/a | n/a | n/a | Most 2025 to 2026 reports start with a live 60-minute screen on HackerRank. Ask your recruiter |
| TikTok (ByteDance) | CodeSignal (most reports naming a platform). HackerRank-style mixed tests | CodeSignal: 4. Americas 2025: 5 MCQs plus 2 coding | CodeSignal 70 min. Americas mixed test 110 min | MCQs. Front-end: a JS/TS debugging task | TikTok sends a preparation kit with sample questions (official FAQ). A 600/600 Sydney candidate got no interview |
| Cloudflare | HackerRank (some tracks) | Not standardized | n/a | none | US intern (Nov 2025): no OA. Canada intern: a HackerRank screen. Some environments make you parse stdin yourself |
| Nutanix | HackerRank | New grad (MTS 1): 2 medium plus CS MCQs (SQL, Linux, DSA) | Not stated | MCQs | Campus drives may apply CGPA and branch cutoffs before the test |
| Airbnb | CodeSignal ICA (Jul to Sep 2026 reports), often sent automatically | 1 task in 4 levels: banking, cloud storage or task manager | 90 min (standard ICA) | none | Screen share and mic stay on. Scores reported: 400, 440, 600 (600 was still rejected). Reports are for SWE and SSE roles, not confirmed for new grad postings |
High-growth tech¶
| Company | Platform | Questions | Time | Extra sections | Notes |
|---|---|---|---|---|---|
| Stripe | HackerRank (most). Unstop for an India intern drive (Sep 2026) | 1 problem in 3 to 6 parts that build on each other | 60 min (some 90) | none | All parts graded by hidden tests (20 in a Jul 2026 report). Reading and typing speed matter more than algorithms. 15 of 17 tests got one intern an interview |
| Databricks | CodeSignal when used, camera on | 3 of rising difficulty (Netherlands, Jan 2025) | Not stated | none | Many 2025 to 2026 intern and new grad candidates had no OA at all |
| Snowflake | HackerRank (most) | 2 to 3, often LeetCode hard (DP, backtracking, graphs) | 90 to 120 min | SQL puzzles or ML questions in some variants | Sent automatically to intern applicants. Copy-paste and tab switching restricted. Official policy (Jan 2026) bans AI on every assessment |
| DoorDash | HackerRank | A few LeetCode medium-hard | Not stated | none | Many US SWE I candidates report no OA. The official FAQ goes from resume review to interviews |
| Lyft | Varies: CoderPad multi-file test, Byteboard, or a timed OA | Timed OA: 2 problems (Canada, Jul 2025). Byteboard: review a design doc, then implement | 90 min or 2 hours | Byteboard design-doc review | Follow the AI rules in your invite |
| Coinbase | CodeSignal plus a vendor cognitive and culture test | Coding: 1 progressive task in 4 levels (some reports: 4 separate problems) | Coding 90 min. Aptitude: about 50 questions in 15 min | Cognitive and culture assessment, about 30 min (official). Personality survey, 100 to 200 items (reported) | Candidates report needing at least 3 of 4 levels. One Apr 2026 loop added an AI-assisted coding OA |
| Robinhood | CodeSignal | Older reports: GCA shape, 4 questions | about 70 min | none | No public cutoff. Thresholds like 820 or 850 in old posts used the retired scale |
| Palantir | HackerRank | Varies: DSA with production-quality code (new grad, Jun 2026); 2 medium plus a REST API task (Jul 2026); FDE intern: one 3-part build (Aug 2026) | FDE intern about 90 min | REST API task in some | In the FDE task, watch rounding rules (half-even to the cent) and keep earlier parts working |
| Roblox | Roblox's own game-based assessment plus CodeSignal | Games (Robots, Factory, Outpost: Mars), about 23 workplace scenarios, 2 coding problems | Coding about 50 min. The whole thing can be spread over a week | Games and a situational test | Practice games Kaiju Cats and Coding Cookies are untimed. One candidate solved both coding questions and was still rejected, so the games count |
| Instacart | CodeSignal | Mid to late 2026: an AI-enabled full-stack repo task in about 5 phases. Earlier 2026: 1 progressive problem in 5 parts, or 4 questions | 90 min (progressive version) | Chat with an AI product manager. One bug fix without AI | Instacart's guide: AI may help organize thoughts but not complete the task, unless the stage asks for AI (Instacart AI usage guide) |
| Rippling | HackerRank | India SDE-1 and campus: 2 DSA, medium to hard. US intern (Jul 2026): 4 MCQs plus 3 coding | 60 min (India) | MCQs (US intern) | Webcam and screen share. Referrals and US experienced candidates often skip the OA |
| Waymo | No standard OA | n/a | n/a | n/a | Live technical screen instead. Official: "the use of AI tools or LLMs is not permitted" unless instructions say so (Waymo how we hire) |
| Anduril | HackerRank | Usually a live 60-minute HackerRank screen. Some roles send a take-home first | 60 min | none | Prompts are vague on purpose. Clarify first, then get a working solution before optimizing |
AI labs¶
| Company | Platform | Questions | Time | Extra sections | Notes |
|---|---|---|---|---|---|
| OpenAI | HackerRank | 2 algorithmic. An invite listed DP, graphs and arrays. Oct 2025 reports: both DP | 110 min (Jan 2025 report) | none | Mainly in the Emerging Talent (intern and new grad) funnel. Camera on and full-screen share. Practice reading stdin |
| Anthropic | CodeSignal | 1 spec in 4 levels. Some 2026 Fellows OAs: extend a mid-sized existing codebase in 5 parts | 90 min | none | Scored out of 600. No AI unless the instructions allow it (Anthropic candidate AI guidance). Topics per Jugal's Fellows post: data structures, algorithms, OOP, small systems, extending code as requirements change |
Finance and quant¶
| Company | Platform | Questions | Time | Extra sections | Notes |
|---|---|---|---|---|---|
| Bloomberg | Usually none for US SWE new grads and interns | n/a | n/a | Plum assessment for certain entry-level roles (official) | Round 1 is usually a live technical interview. Some roles use pre-recorded video or skills tests. The recruiter tells you |
| Goldman Sachs | HackerRank | Official (Aug 2026): Software Fundamentals (DSA, any language, everyone). Software Diagnostics (Engineering roles). Math MCQs (Quantitative Strategist roles only) | 30 min, 45 min, 22 min (official) | Diagnostics: diagnose and fix a Django, Node.js or Spring Boot app with a "limited" AI assistant that will not give the full solution | 2025 to 2026 reports describe an older combined coding plus math test stated as 180 min: varies by role and cycle, follow your invite. Update "within three weeks" (Goldman Sachs HackerRank guide). Campus candidates report a HireVue after the OA |
| JPMorgan Chase | HackerRank, video proctored (2026 reports) | 2 coding | about 60 min | On-demand recorded video interview (official). Some India tests add MCQs | Hidden tests are strict: brute force passes samples, fails hidden cases. Official: "many assessments may not be retaken for a designated timeframe", and JPMorgan recommends no more than three summer internship applications (JPMorgan FAQ) |
| Capital One | CodeSignal, proctored | 4 | 70 min | none | A matrix-command simulation (swap rows, rotate) recurs. 600/600 does not guarantee a Power Day. At most 5 active US applications at once (official FAQ) |
| Morgan Stanley | SHL/AMCAT (India campus). HackerRank (lateral; US per secondary sources) | India campus: 7 debugging, 24 interactive aptitude, 3 coding | about 111 min (15 + 36 + 60) | Debugging (save your code) and interactive aptitude | Coding questions are reworded LeetCode classics (House Robber, Delete and Earn) |
| Citadel | HackerRank | Campus SWE: 2 to 3, medium to hard. Citadel Securities C++: 1 performance refactor. NXT program: 14 MCQs plus 1 coding | 60 to 75 min (one 2025 invite: 66 min). NXT: 90 min | NXT MCQs | Official campus page starts with a video interview, but OA reports continue through Summer 2026. Expect DP and counting mod 1e9+7 |
| Two Sigma | HackerRank | 2, medium to hard | 90 min (2024 report) | none | Postings say AI tools are strictly prohibited. One intern solved 1 of 2 and still advanced |
| Jane Street | Usually none for US SWE | n/a | n/a | Trading roles: probability tests | HackerRank tests reported for first and second-year programs and Asia campus hiring. Prepare for a live screen (Jane Street interviewing) |
| D. E. Shaw | HackerRank (most) | India new grad: 3 coding, 10 technical MCQs, 10 aptitude MCQs. US (Oct 2025): 4 medium-hard | India 120 min. Senior 90 min | MCQs with -0.5 for wrong answers. Sections cannot be revisited | 1 full plus 1 partial solution has advanced |
| Hudson River Trading | HackerRank or Codility (official 2021 blog) | 2025 intern reports: 4 easy to medium with many hidden tests | Timed | none | Books and the internet are allowed as a language reference (2021 blog). 2026 postings say AI tools are strictly prohibited unless instructed |
India and Asia¶
| Company | Platform | Questions | Time | Extra sections | Notes |
|---|---|---|---|---|---|
| Flipkart | HackerEarth (lateral). GRiD 8.0 registration on Flipkart's own event page (earlier GRiD editions used Unstop) | Freshers: 3 DSA (easy, medium, hard). GRiD 8.0 (Jul 2026): 2 DSA, 1 SQL, 1 Python DataFrame | Freshers 60 to 90 min. GRiD 8.0: 120 min reported (the event page said a 90-minute test): varies | SQL and pandas (GRiD) | Long story-style problems with STDIN/STDOUT. Shortlists go by test cases passed (GRiD 8.0 page) |
| PhonePe | CodeSignal (lateral). Campus platform not named, camera on | 4, DP-heavy, Codeforces-style | Lateral about 70 min | none | 2 full plus 2 partial solutions got one candidate shortlisted |
| Agoda | HackerRank (official) | Interns: 3 to 4 DSA. IC1 (Jul 2025): 3. Experienced: plus a REST API question | 30 to 60 min (official guide). IC1: 75 min | REST API (experienced) | Agoda's guide recommends a practice round and a stable connection |
If your company is not here¶
- Check more companies for question lists.
- Search the company's LeetCode Discuss topic for "OA" and read only the last 3 months.
- Read the platform section on OA platforms. Most OAs are one of the three big platforms.
- Ask the recruiter. Use this message:
Hi [Recruiter name],
Thank you for the assessment link for [Role, Job ID]. Before I start, could you confirm:
1. Which programming languages can I use?
2. Can I use official language documentation during the test?
3. Is any AI assistant allowed, or none at all?
4. Is the test proctored (webcam, screen share, ID)?
Thank you,
[Your name]