Rotting Oranges (LeetCode 994) — You are given an m x n grid where each cell is either 0 (empty), 1 (fresh orange), or 2 (rotten orange). Every minute, any fresh orange 4-directionally adjacent to a rotten orange becomes rotten. Return the minimum number of minutes until no fresh orange remains, or -1 if it is impossible.

The key insight is multi-source BFS starting simultaneously from all rotten oranges at time 0. Enqueue all cells with value 2, then BFS layer by layer, each layer representing one minute. Count fresh oranges up front; decrement as they rot and return layer count if all are consumed. Time O(m*n), space O(m*n). Follow-up: what if oranges rot diagonally too — just add 4 more neighbors. Ability to recognize multi-source BFS vs single-source; correctly tracking remaining fresh count; handling the -1 (isolated fresh) case cleanly.

Minimum Number of Arrows to Burst Balloons (LeetCode 452) — There are spherical balloons taped onto a flat wall. Each balloon is represented as [x_start, x_end]. An arrow shot vertically at x bursts every balloon with x_start <= x <= x_end. Find the minimum number of arrows.

Greedy interval scheduling: sort balloons by end coordinate. Shoot the first arrow at the end of the first balloon; skip all balloons whose start falls within that shot; when a balloon is missed, fire a new arrow at its end. O(n log n) sort + O(n) scan. Follow-up: what if arrows cost differently — becomes weighted set cover, NP-hard. Recognizing the greedy interval-merging pattern and why sorting by end (not start) is correct; clean boundary handling.

Reverse Nodes in k-Group with alternating group sizes — Given a linked list and two integers x and y, reverse the first x nodes, leave the next y nodes as-is, and repeat. (Variant of LeetCode 25 Reverse Nodes in k-Group.)

Generalize k-group reversal: track a phase toggle between reversing x nodes and skipping y nodes. For each group, verify enough nodes exist before reversing; use a prev pointer to splice the reversed segment back. O(n) time, O(1) space. Follow-up: what if x equals y — every other group, simplifies but logic stays the same. In-place linked list pointer manipulation; handling tail of list where fewer than x nodes remain; generalizing cleanly from the pure k-group problem.

Design an Inventory Management System for an e-commerce platform (admin side: add/update/remove products, track stock levels, handle concurrent purchases decrementing stock).

Start with requirements: read-heavy product catalog, write-heavy stock decrements on purchase. Use a relational DB (products and inventory tables) with row-level locking or optimistic concurrency (version column) for stock updates. Cache product metadata in Redis; use a message queue (Kafka) to decouple order events from inventory updates. For flash sales, reserve stock in Redis atomically with DECR and reconcile with DB async. Follow-up: handle oversell during a spike. Clear separation of catalog (read path) vs inventory mutation (write path); knowledge of race conditions and how optimistic locking or queue-based processing resolves them; awareness of cache invalidation.

Design a Notification Service that sends notifications across multiple channels (email, SMS, WhatsApp, push notifications).

Producer side: notification events published to a topic per channel on Kafka. Per-channel consumer services poll and call the respective provider API (SendGrid, Twilio, FCM). Store notification records in DB for dedup (idempotency key = user+event+timestamp). Retry with exponential backoff; dead-letter queue for permanent failures. Rate-limit per user per channel. Follow-up: how do you guarantee exactly-once delivery vs at-least-once. Observer/pub-sub pattern recognition; channel abstraction via strategy pattern in the LLD layer; dedup and retry correctness; understanding of provider-level rate limits.

Design a URL Shortener service (like bit.ly): generate short aliases for long URLs and redirect users to the original URL.

Hash long URL to a 6-7 character base-62 string (use a counter + base-62 encode or MD5 and take first 7 chars, handling collisions). Store {shortCode -> longUrl} in a DB; cache hot entries in Redis with TTL. On redirect, read from cache or DB and issue 301 (permanent) or 302 (trackable) redirect. For analytics, log redirect events asynchronously. Scale reads with replicas; scale writes with a distributed ID generator (Snowflake). Follow-up: custom aliases, expiration, analytics dashboard. Base-62 encoding rationale; cache-aside pattern; 301 vs 302 tradeoff for analytics; collision handling strategy.

Tell me about a time you used data or metrics to improve or refine a process. What did you measure, what did you change, and what was the outcome?

Structure with STAR; tie to Adobe's 'Own the Outcome' value. Quantify baseline metric, intervention, and result (e.g., reduced build time by 40%). Be specific about which metric you chose and why, and acknowledge any metric that did not improve. Interviewer will probe: how did you decide that metric was the right one to track? Data-driven decision making; ownership without being assigned the work; ability to quantify impact.

Describe a time you went above and beyond for a customer or stakeholder. What motivated you and what did you do that was outside your normal scope?

This maps to Adobe's 'Genuine' and 'Adobe for All' values. Frame the extra action as customer-empathy-driven rather than self-promotion. Include what you learned about the customer's real need, what you did, and how it changed the relationship or product. Follow-up: how did you balance this extra effort against your sprint commitments? Genuine customer orientation; willingness to cross functional boundaries; judgment about when going above and beyond is the right call vs. scope creep.

Tell me about an idea you came up with and implemented. What was the impact, and what would you do differently?

Maps to 'Create the Future' (Innovation). Describe the problem you observed proactively (not assigned), the idea, how you got buy-in, the execution, and a measurable result. The 'what would you do differently' component is critical — interviewers want intellectual honesty. Follow-up: how did you handle pushback from teammates who preferred the existing approach? Proactive problem identification vs waiting to be assigned; ability to influence without authority; honest reflection — candidates who say 'I would not change anything' score poorly.

Adobe Interview Questions

Q: Minimum Number of Arrows to Burst Balloons (LeetCode 452) — There are spherical balloons taped onto a flat wall. Each balloon is represented as [x_start, x_end]. An arrow shot vertically at x bursts every balloon with x_start <= x <= x_end. Find the minimum number of arrows.

Greedy interval scheduling: sort balloons by end coordinate. Shoot the first arrow at the end of the first balloon; skip all balloons whose start falls within that shot; when a balloon is missed, fire a new arrow at its end. O(n log n) sort + O(n) scan. Follow-up: what if arrows cost differently — becomes weighted set cover, NP-hard. Recognizing the greedy interval-merging pattern and why sorting by end (not start) is correct; clean boundary handling.

Q: Reverse Nodes in k-Group with alternating group sizes — Given a linked list and two integers x and y, reverse the first x nodes, leave the next y nodes as-is, and repeat. (Variant of LeetCode 25 Reverse Nodes in k-Group.)

Generalize k-group reversal: track a phase toggle between reversing x nodes and skipping y nodes. For each group, verify enough nodes exist before reversing; use a prev pointer to splice the reversed segment back. O(n) time, O(1) space. Follow-up: what if x equals y — every other group, simplifies but logic stays the same. In-place linked list pointer manipulation; handling tail of list where fewer than x nodes remain; generalizing cleanly from the pure k-group problem.

Q: Design an Inventory Management System for an e-commerce platform (admin side: add/update/remove products, track stock levels, handle concurrent purchases decrementing stock).

Start with requirements: read-heavy product catalog, write-heavy stock decrements on purchase. Use a relational DB (products and inventory tables) with row-level locking or optimistic concurrency (version column) for stock updates. Cache product metadata in Redis; use a message queue (Kafka) to decouple order events from inventory updates. For flash sales, reserve stock in Redis atomically with DECR and reconcile with DB async. Follow-up: handle oversell during a spike. Clear separation of catalog (read path) vs inventory mutation (write path); knowledge of race conditions and how optimistic locking or queue-based processing resolves them; awareness of cache invalidation.

Q: Design a Notification Service that sends notifications across multiple channels (email, SMS, WhatsApp, push notifications).

Producer side: notification events published to a topic per channel on Kafka. Per-channel consumer services poll and call the respective provider API (SendGrid, Twilio, FCM). Store notification records in DB for dedup (idempotency key = user+event+timestamp). Retry with exponential backoff; dead-letter queue for permanent failures. Rate-limit per user per channel. Follow-up: how do you guarantee exactly-once delivery vs at-least-once. Observer/pub-sub pattern recognition; channel abstraction via strategy pattern in the LLD layer; dedup and retry correctness; understanding of provider-level rate limits.

Q: Design a URL Shortener service (like bit.ly): generate short aliases for long URLs and redirect users to the original URL.

Hash long URL to a 6-7 character base-62 string (use a counter + base-62 encode or MD5 and take first 7 chars, handling collisions). Store {shortCode -> longUrl} in a DB; cache hot entries in Redis with TTL. On redirect, read from cache or DB and issue 301 (permanent) or 302 (trackable) redirect. For analytics, log redirect events asynchronously. Scale reads with replicas; scale writes with a distributed ID generator (Snowflake). Follow-up: custom aliases, expiration, analytics dashboard. Base-62 encoding rationale; cache-aside pattern; 301 vs 302 tradeoff for analytics; collision handling strategy.

Adobe's SWE loop (titled "Computer Scientist 1/2" or "MTS-1/2" in India and "Software Engineer" globally) runs 3-5 rounds: an online assessment (CodeSignal or HackerRank, 2-3 problems), 2-3 DSA coding rounds on an IDE (each ~60 min, typically C++ or Java), a system design round (HLD + LLD for mid/senior), and a hiring-manager or director behavioral close. New-grad and intern loops are lighter (OA + 1-2 technical rounds + 1 behavioral). The process has been predominantly virtual since 2022 with occasional on-site final rounds for senior India roles.

14 real questions across 3 rounds

Reported via LeetCode Discuss first-hand reports: Adobe Computer Scientist Interview Experience ALL Rounds (post 6148547, Dec 2024 offer) — explicitly names Rotting Oranges and Minimum Arrows; Adobe interview experience Noida Sr. Software Engineer offer accepted (post 5784907) — explicitly names the alternating x-y reverse nodes variant; Adobe CS-1 interview experience (post 6943130) — reports Container With Most Water

Run a Adobe mock with the AI interviewer. It asks questions like these, follows up on your answers, and tells you exactly what to fix. The only score that means anything is the one on your real answers.

Continue with Google

Keep prepping

All company interviews