The Coding Interview

The problem Oracle reaches for more than any other is LRU Cache, a hash map wired to a doubly linked list. It turns up about three times as often as the next problem in the pile. Start there and you have already read the room. Oracle's coding bar sits in the easy-to-medium band, and medium is the mode. A candidate who is fluent at medium arrays, strings, trees, and stacks clears most Oracle loops, and the hard problems that do appear lean hard-looking rather than exotic. Dynamic programming is here too, moderate and not banned, so do not skip it.

The Oracle angle

Ask the recruiter which door you're walking through. It decides what to prep. The campus assessment rewards breadth: fast arithmetic, database and operating-systems trivia, output-prediction on trees, and one clean coding solve under a tight clock. The experienced loop rewards depth: fewer questions, more follow-up, complete working code you defend line by line. A lateral hire who spent a month on aptitude drills prepped the wrong door, and so did a new grad who skipped the DBMS multiple choice. The content overlaps, the format does not, and Oracle rarely tells you which one is coming unless you ask.

The Two Front Doors

Two paths lead into the same coding content, and the recruiter decides which one you take. Campus and new-grad hiring, heaviest in India, runs through a HackerRank assessment. Experienced and lateral hiring skips it and goes straight to a live loop. The questions rhyme across both doors. The wrapper does not.

• Campus / new-grad OA: HackerRank, ~100 min, ~36 questions. The long front door. Roughly fifteen aptitude questions, about twenty technical multiple-choice questions on databases, operating systems, and object-oriented concepts, then one moderate coding problem. A common variant runs three coding problems instead, one algorithm plus one SQL plus one small API task. The database-company fingerprint is right there in the assessment.
• Experienced / lateral loop: No OA, then 4 to 5 live rounds. A phone screen followed by four or five one-hour rounds on an online compiler. Rounds mix medium algorithmic coding, low-level or object-oriented design, and behavioral, with a lighter system-design discussion added at senior levels only. Java is the house language since Oracle owns it, but you code in whatever you like.
• The shared bar: Easy to medium, medium is the mode. Whichever door you take, the coding content lands in the same band. Complete working code is the gate, not a clever one-liner. Hard problems are a minority and skew hard-looking rather than algorithmically exotic, so a strong medium solver reasons through them.

The Oracle angle

Oracle wants the algorithm instead of the library call, and the code has to run. In the live loop you write on an online compiler, and your solution has to compile and pass the test cases. A partial pass is recoverable. Built-in helpers are sometimes taken away on purpose: skip the language's sort, do it in place, find something without the square-root helper. So bring your own test cases, name your edge cases before you are asked (the empty input, the duplicate key, the out-of-range value), and show the brute force before you optimize. Clean, complete, compiling code beats a clever fragment that does not run.

The Question Mix

The cluster weights below reflect how often each family shows up in Oracle rounds. Treat them as directional rather than published: Oracle does not release this, and the point is the shape, not the decimal. Arrays and strings lead, as they do almost everywhere. But two clusters punch above their raw share at Oracle specifically. Design and LRU is one. The interval-and-scheduling sweep is the other.

Read the chart as a map of where your first hours go, not a leaderboard. Arrays and hashing is the workhorse, and it carries the campus assessment. Strings and parsing sits right behind it, heavy on sliding-window work and the fiddly decode-and-format problems Oracle likes. Graphs, grids, and dynamic programming are grouped because they travel together in the loop. That's where the moderate DP tier lives. Two Premium questions are fair game, so the old line that Oracle never asks DP is wrong. Trees lean toward building and validating a BST rather than generic traversal. Then come the two Oracle-favored clusters that punch above their share, Design and LRU and the interval sweep.

The Oracle angle

Design and LRU Cache is the strongest single signal in the Oracle loop. Over-prepare it. LRU Cache alone shows up about three times as often as the next problem. A Premium question in O(1) rides alongside it. The reason it counts twice: it sits on the seam between a coding problem and a low-level design problem. The same hash-map-plus-doubly-linked-list build that clears a coding round also anchors the machine-coding round at senior levels, sometimes with a thread-safe twist. Get LRU cold, forwards and backwards, and you have covered the most likely single hour of your loop.

The Fundamentals Layer

A generic algorithm playbook stops here, and Oracle keeps going. Because Oracle builds databases, a real CS-fundamentals layer sits next to the coding. Database theory shows up directly (normalization, ACID, keys, indexing), alongside operating-systems and object-oriented multiple choice and output-prediction on trees, heaviest in the campus assessment's technical block. SQL is genuine, from joins and group-by to subqueries, and it goes deeper into PL/SQL for database-leaning roles. Phone screens sometimes probe Java internals: how a HashMap works under the hood, collections, thread pools, volatile versus atomic. At senior levels a distinct concurrency round appears, with producer-consumer, a thread-safe cache, a singleton that survives threads. A candidate can ace every array problem and still stall here.

The Oracle angle

Oracle is a database company. Prep only algorithms and you walk in blind to half the screen. The campus assessment devotes a whole multiple-choice block to databases, operating systems, and object-oriented concepts. It's not unusual for the coding section to bundle a SQL problem next to the algorithm. In the experienced loop the same layer surfaces as SQL-versus-NoSQL and indexing trade-offs inside a design round, and some rounds even let you choose database queries or algorithms. None of that is on a generic problem list. A pure grinder misses this part, so the Oracle Fundamentals lesson covers it. That's where you work a self-join over a manager hierarchy and a GROUP BY with a HAVING aggregate.

The Lessons Ahead

Seven lessons follow. Six hold the DSA spine, and the seventh carries the Oracle-specific fundamentals. They open with the arrays-and-hashing workhorse and move through strings, trees, the design and LRU cluster, stacks and intervals, and the combined graphs, grids, and DP band. Then comes the Oracle Fundamentals lesson that decides rounds a pure algorithm candidate never saw coming. Each page opens with how Oracle frames the family, then works the recurring problems in the compile-it-and-defend-it style the interviewer expects. Where a lesson rests on a base algorithm, the full build lives in the free Foundations course, so these pages stay on what is specifically Oracle.

Common questions