115 questions from 14 interviews · updated from reports 2014-2024
About
Amadeus is a travel technology company that builds reservation, distribution, and IT systems used by airlines, hotels, and travel agencies. In India, it is known for hiring Software Engineers, Software Developers, and SE1 roles for software development, testing, and support.
The roles that come up most are Software Engineer, Software Developer and SE1. This covers 14 candidate interviews reported from 2014 to 2024. Most sat it at entry level (13 of 14 that recorded a level), with 1 internship interviews alongside. Among the 14 that recorded either route, arrivals split between campus drives (14, 100%). Most questions fall under CS fundamentals and DSA.
Ans. Kadane’s Algorithm finds the maximum sum of a contiguous subarray in linear time. It scans the array while keeping the best sum ending at the current position and the best sum seen overall. At each element, either extend the previous subarray or start a new one. It runs in O(n) time and O(1) space.
Ans. Dijkstra’s algorithm finds the shortest path from a source node to all other nodes in a weighted graph with non-negative edge weights. It keeps tentative distances, repeatedly picks the unvisited node with the smallest distance using a priority queue, and relaxes its neighbours. With an adjacency list, it runs in O((V + E) log V).
Ans. Maps are usually implemented as either hash tables or balanced search trees. A hash map stores key value pairs in buckets chosen by a hash function, handling collisions by chaining or probing, giving average constant time lookup. Tree maps keep keys ordered, often with a red black tree, giving logarithmic operations.
Ans. Cache memory is a small, very fast memory between the CPU and main memory that stores recently or frequently used data and instructions. It is implemented with SRAM in levels such as L1, L2 and L3, using cache lines, tags, valid bits, replacement policies, and write policies to manage access efficiently.
Ans. Graphs can be directed or undirected, weighted or unweighted, cyclic or acyclic, connected or disconnected, and sparse or dense. A tree is a connected acyclic graph, commonly rooted, binary, binary search, balanced, heap, trie, or B-tree. The key distinction is that graphs are general relationships, while trees enforce hierarchy without cycles.
Ans. Use singly linked nodes and keep the right pointers for each structure. For a stack, keep a top pointer; push inserts at the head and pop removes from the head. For a queue, keep front and rear pointers; enqueue adds at rear and dequeue removes from front. All main operations are O(1).
Ans. A linked list is a linear data structure where each node stores a value and a reference to the next node, and sometimes the previous node. The key trade-off is fast insertion or deletion when you already have the node, but slow random access because you must traverse from the head. Common questions use two pointers.
Ans. Design it as gate controllers, sensors, payment terminals and a central service that tracks spaces, tickets, pricing and access. Cars receive a ticket or are recognised by number plate, spaces are allocated by type, occupancy updates come from sensors, and exit is allowed after payment. The key detail is keeping occupancy state strongly consistent.
Ans. Merge sort is a divide and conquer sorting algorithm that splits the array into halves, sorts each half recursively, then merges the sorted halves. The key step is merging by comparing the smallest remaining elements. It runs in O(n log n) time and usually needs O(n) extra space.
Ans. Semaphores are synchronisation primitives used to control access to shared resources, and a critical section is the part of a program that accesses shared data and must not be executed by multiple conflicting threads at once. A semaphore is changed using wait and signal operations, commonly to enforce mutual exclusion or limit concurrency.
Ans. Implement it by wrapping a HashMap inside the collection class and exposing methods such as add, remove, contains and size. Store each element as a key and a dummy value, or store keys to counts if duplicates are allowed. Add, remove and lookup are average constant time, with linear space.
Ans. Processes are independent running instances of programs, each with its own virtual address space, resources and execution state. They communicate using inter-process communication such as pipes, sockets, message queues, signals, files or shared memory. The key detail is isolation: normal memory is not shared, so shared memory needs explicit synchronisation such as locks or semaphores.
Ans. Read the table, chart, or graph carefully, noting units, totals, and time periods. Identify exactly what is being asked, then extract only the relevant figures. Use standard calculations such as percentage change, ratios, averages, or differences. Estimate first to avoid impossible answers, then calculate accurately and check units before choosing the answer.
Ans. Inter-process communication is accomplished through operating system mechanisms that let separate processes exchange data and coordinate actions. Common methods include pipes, message queues, shared memory, sockets, signals, and files. The key detail is synchronisation: shared memory is fast, but processes must use locks, semaphores, or similar controls to avoid races.
Ans. An application is slow because some resource or dependency is the bottleneck. Common reasons include inefficient algorithms, slow database queries, missing indexes, excessive network calls, high CPU or memory use, disk I/O, lock contention, too much logging, poor caching, or overloaded servers. The key is to measure with profiling and monitoring, not guess.
Ans. Model the ER diagram with Student, Course, Teacher, ClassSession and Attendance as the main entities. Student enrols in Course, Teacher teaches Course, Course has many ClassSessions, and Attendance links one Student to one ClassSession with status, time and remarks. The key detail is making Attendance an associative entity to support many-to-many tracking.
Ans. Naive Bayes is a probabilistic classifier using Bayes’ theorem with an assumption that features are independent. Support Vector Machine is a classifier that finds the best separating boundary by maximising the margin between classes. The key detail is that Naive Bayes is fast and simple, while SVMs handle high-dimensional data well, especially with kernels.
Ans. The STL is C++’s generic library of containers, algorithms, iterators and function objects. A std::list is usually implemented as a doubly linked list: each node stores a value plus pointers to previous and next nodes. Insert and erase are constant time with an iterator, but indexing and search are linear.
Ans. Use one weighing test, assuming contaminated pills have a known different weight. Number the jars 1 to n. Take 1 pill from jar 1, 2 from jar 2, and so on, then weigh all sampled pills together. The excess or shortfall from the expected weight identifies the jar number directly.
Ans. Use Dijkstra by keeping the best known distance to each vertex, starting with the source at 0, and repeatedly expanding the unvisited vertex with the smallest distance. Store adjacency lists for the graph and use a min-priority queue. Relax each outgoing edge when a vertex is popped. Time complexity is O((V + E) log V).
Ans. Use dynamic programming with a 2D table where dp[i][j] stores the LCS length for the first i characters of one string and first j of the other. If characters match, add one from dp[i-1][j-1]; otherwise take the maximum of top or left. Time and space are O(nm).
Ans. Scan every cell, and when you find unvisited land, count one island and traverse all connected land using DFS or BFS. DFS uses recursion or a stack, while BFS uses a queue. Mark visited cells to avoid recounting. Check four directions unless diagonals are specified. Time is O(rows × cols), space is O(rows × cols).
Ans. I would identify duplicates using the columns that define uniqueness, keep one row from each group, and delete the rest. In SQL, this is commonly done with a window function such as row number over the duplicate key. The important detail is to run it in a transaction, verify first, then add a unique constraint to prevent recurrence.
Ans. Implement a binary tree using nodes that store a value, a left child reference, and a right child reference, then perform in-order traversal by visiting left subtree, current node, then right subtree. This is usually done recursively with a base case for null nodes. Time complexity is O(n), and space is O(h).
Ans. Read the questions first, then scan the passage for the relevant lines. Base every answer only on what the passage states or clearly implies, not outside knowledge. Note keywords, tone, comparisons and conclusions. For inference questions, choose the option most directly supported. Eliminate choices that are extreme, unrelated or only partly true.
Ans. Multiprogramming runs several programs by sharing one CPU, multiprocessing uses multiple CPUs or cores to run work in parallel, and multithreading runs multiple threads within one process. The key difference is the unit and hardware involved: programs share CPU time, processes can run on separate cores, and threads share the same process memory.
Ans. I would implement a dictionary as a hash table: store key value pairs in an array of buckets, compute a hash from the key, and map it to a bucket index. Handle collisions with chaining or open addressing. Keep the load factor bounded by resizing. Average lookup, insert, and delete are constant time.
Ans. Put one white ball alone in one box, and put all remaining balls in the other box. If there are 50 white and 50 black balls, the chance is 1/2 times 1 plus 1/2 times 49/99, which is about 74.7%. This works because one box gives certain success while the other remains nearly half white.
Ans. Convert the wording into numbers, identify what is being asked, and choose the relevant operation or formula. Work step by step, keeping units consistent. Estimate first to spot unreasonable answers, then calculate accurately. For percentages, ratios, rates, averages, and tables, write the relationship clearly before solving. Double-check the final value against the question.
Ans. Read the passage or statement carefully, then base every answer only on the information given, not outside knowledge. Identify key facts, qualifiers, negatives, and comparisons. For logic questions, translate statements into simple relationships and test each option. Eliminate answers that are too broad, unsupported, or contradict the text.
Ans. Choose a situation involving a distressed customer, financial risk, and clear process. Emphasise empathy, identity verification, immediate card or account protection, accurate dispute logging, escalation to fraud teams, regulatory timelines, and follow-up. Interviewers listen for calm judgement, customer care, attention to detail, confidentiality, and balancing speed with compliance.
Ans. Read the title, units, scale, and labels first. Identify exactly what is being asked, then extract only the needed values from the graph, chart, or table. Use ratios, percentages, averages, differences, or totals as required. Keep calculations organised, watch for changing units, and check whether the answer should be approximate or exact.
Ans. Common searching algorithms are linear search, which checks each item in O(n), and binary search, which halves a sorted array in O(log n). Common sorting algorithms include bubble, selection and insertion sort in O(n²), merge sort and heap sort in O(n log n), and quicksort, usually O(n log n) but worst-case O(n²).
Ans. Implement it with a Trie where each node stores child links and a flag marking the end of a valid word. Insert each word character by character, search by following characters, and support prefix lookup by stopping at the prefix node. Insert and search take O(L) time, where L is word length.
Ans. The diamond problem occurs when a class inherits from two classes that both inherit from the same base, creating two base subobjects and ambiguity. Resolve it in C++ with virtual inheritance: make both intermediate classes virtually inherit the base. No special data structure or algorithm is involved, and runtime complexity is unchanged.
Ans. Estimate wall area first. If the block is 30 m by 20 m, 4 floors high, height 12 m, exterior area is about 1,200 m² after openings. With two coats at 8 m² per litre, paint needed is 300 litres. At £5 per litre plus labour, total is about £3,000 to £5,000 and 5 to 7 days.
Ans. MongoDB is a document-oriented NoSQL database, while relational databases store data in tables with fixed schemas and relationships. MongoDB stores JSON-like BSON documents, so related data is often embedded together and schemas can vary between records. This makes it flexible, but data modelling and joins differ from SQL-based systems.
Ans. Read the chart or table title, units, labels, and time period first. Identify exactly what is being asked, then pick only the relevant figures. Use simple arithmetic such as totals, differences, averages, percentages, ratios, or growth rates. Estimate where possible, but calculate accurately when options are close. Check units and rounding before answering.
Ans. Use an event driven system that ingests signals from mobile networks, transport gates, road cameras, and official check-ins, then maintains a real time estimated population per city. The key detail is avoiding double counting: assign probabilistic anonymous device or person IDs, deduplicate within time windows, and publish counts with confidence intervals rather than exact numbers.
Ans. The final colour is determined by the parity of the initial number of white marbles, for the common rule where two different colours return a white marble and two same colours return a black marble. Each move preserves whether the number of white marbles is odd or even. Odd means final white, even means final black.
Ans. Short term, I would stabilise the database by reducing load, killing runaway queries, freeing temporary space, increasing memory if possible, and restarting only if safe. Long term, I would find the cause using metrics, optimise queries and indexes, tune memory settings, archive old data, fix leaks, and plan capacity before demand grows.
Ans. Convert all figures to a common base first, usually cost price or total value. For profit and loss, use profit percent equals profit divided by cost price times 100. For percentages, translate words into equations. In data interpretation, read units carefully, compare like with like, and calculate ratios, differences, or percentage change as required.
Ans. The String Pool is a special area in the Java heap that stores unique String literals so identical values can share the same object. When a literal is created, Java reuses an existing pooled String if present. Using new String creates a separate object, while intern() can add or return the pooled instance.
Ans. Collisions in a HashMap are handled by storing multiple entries in the same bucket, usually as a linked list or, in Java, a tree when collisions grow large. A key’s hash is converted to a bucket index. Within that bucket, keys are compared using equals to find, update, or insert the correct entry.
Ans. Reduce the fraction to its simplest form. Then check the denominator. If its only prime factors are 2 and 5, the decimal terminates. If any other prime factor remains, it is recurring. This works because finite decimals are fractions whose denominators can divide a power of 10.
Ans. Read the passage carefully and base every answer only on what it states or clearly implies. Identify key facts, qualifiers, comparisons and cause-effect links. For each option, check whether it is true, false, or cannot be determined from the text. Do not use outside knowledge or assumptions, even if they seem reasonable.
Ans. The engine stops after the largest time t for which cumulative power used is still within the available power. Compute the running sum of consumption per time unit and stop when it first exceeds the capacity. No special data structure is needed. The time complexity is O(n) and space is O(1).
Ans. Read the table or graph carefully, noting units, scales, and categories. Identify exactly what is being compared. For ratios, put values in the same units and simplify. For percentages, use part divided by whole times 100. For trends, compare changes over time and note increases, decreases, peaks, and exceptions.
Ans. Resolve the initial velocity into horizontal and vertical components: ux = u cos θ and uy = u sin θ. Use horizontal motion x = ux t and vertical motion y = uy t minus half gt squared. Choose the equation containing the unknown, apply given conditions, then solve algebraically, using g ≈ 9.8 m/s².
Ans. Arrays give fast indexed access, linked lists make insertions easier, stacks and queues control order, hash tables give fast key lookup, trees keep data ordered, heaps find priorities, and graphs model relationships. Choose by access pattern: random access, frequent updates, lookup by key, sorted traversal, priority retrieval, or connected entities.
Ans. Base every answer only on the paragraph, not on outside knowledge or assumptions. Mark True if the statement is directly supported, False if it directly contradicts the text, and Cannot Say if the paragraph does not give enough information. Read qualifiers carefully, such as all, some, always, before, after, and may.
Ans. Implement them with classes: inheritance by subclassing, polymorphism by overriding methods and calling the same interface, abstraction by defining essential behaviour with base or abstract classes, and encapsulation by keeping state inside objects with controlled access. The key detail is designing clear interfaces; runtime cost is usually just normal method dispatch.
Ans. Read the passage carefully and treat only its stated facts as true. For each statement, check whether it is directly supported, directly contradicted, or cannot be decided from the text. Do not use outside knowledge or assumptions. Watch for qualifiers such as all, some, usually, never, and compare wording exactly.
Ans. Trace the code manually, not by guessing the language. Identify initial values, loop bounds, update steps, and operator precedence. Build a small table for changing variables. Watch for integer division, post and pre-increment, scope, and off-by-one errors. Follow the given pseudocode rules consistently until the final print statement.
Ans. Decorators wrap functions or classes to add behaviour; range creates sequences in Python 3 while xrange was Python 2’s lazy iterator. Python passes object references by assignment, so mutables can be changed. __init__ initialises new objects. Namespaces map names to objects. *args collects extra positional arguments, **kwargs collects extra named arguments.
Ans. I would first boot it from a USB drive with a Linux installer or live environment, then install an OS and development tools. I would choose C initially, because it is close to the hardware, works well with minimal runtime support, and helps if I need to write or understand low-level system code.
Ans. OOP uses encapsulation, abstraction, inheritance and polymorphism to model real entities as objects. In a payment system, Payment hides card details, exposes pay, and CardPayment inherits it. Overloading means pay(amount) and pay(amount, currency) in one class. Overriding means CardPayment changes the parent pay behaviour. Method choice is constant time.
Ans. Model a car system with a Car class encapsulating speed and fuel, a Vehicle base class for inheritance, a start method overridden by ElectricCar for polymorphism, and an abstract Serviceable interface for abstraction. Use objects holding fields and method tables. Operations like start, accelerate, and refuel are O(1).
Ans. Pick an example where you set direction, clarified roles, supported people, and handled a problem without taking over. Emphasise communication, trust, accountability, feedback, and adapting your style to different team members. Interviewers listen for calm leadership, fairness, measurable outcomes, and evidence that you develop people rather than just control tasks.
Ans. Choose a real example where the conflict affected delivery, not just personalities. Emphasise that you listened to both sides, clarified facts, kept focus on the shared goal, and helped agree a practical next step. Interviewers listen for maturity, calm communication, accountability, and a positive outcome or lesson learned.
No question matches those filters. Reset filters.
Showing 60 of 115 questions. Ranked by how often the same question came back across interviews.
A spoken interview built from these questions, scored when you finish; the feedback is yours.
Start practisingWhen you are ready, record The One: a single interview hiring teams watch, so you stop repeating first rounds.
Candidate interviews most often cover CS fundamentals (46%) and DSA (21%).
Candidate interviews show an average of 3.7 rounds per experience, with a typical sequence of Online test → Online test → Technical. Individual interview paths can vary.
Among questions with a recorded difficulty, the mix is easy 40%, medium 58%, hard 2%.
