Interview experience
Akamai Interview Questions and Answers (2026)
Overview
Section titled “Overview”Akamai’s fresher loop looks like a standard product-company process on paper, but its technical and managerial rounds lean unusually hard on networking and security depth given its business as a global CDN and edge-security provider.
Akamai interview process at a glance
Section titled “Akamai interview process at a glance”| Round | Duration | What they test |
|---|---|---|
| Aptitude test (some drives) | ~30 min | General reasoning |
| Online Assessment (HackerRank) | 60-75 min | CS fundamentals MCQs (OS, DBMS, networks, OOPs), 1 SQL question, 2 coding problems |
| Group discussion (some drives) | ~15 min | Communication, articulation on networking/security topics |
| Technical Interview(s) | 45-60 min | DSA, coding, OOP concepts, deep Computer Networks and IT Security |
| Managerial/HR Round | 20-45 min | Linux/OS/CN/DBMS review, fit, offer discussion |
Online Assessment
Section titled “Online Assessment”A HackerRank paper run around 75 minutes: roughly 9 CS-fundamentals MCQs (OS, CN, DBMS, OOPs, DS), one SQL query, and two coding problems on array manipulation and string processing.
Common questions
- Array manipulation and string-processing coding problems
- SQL query involving a join or aggregate function
- MCQs on OS scheduling, DBMS normalization, and OOP concepts
Group discussion (some drives)
Section titled “Group discussion (some drives)”On drives that include it, candidates are split into groups of around 5 and asked to discuss a topic related to Akamai’s domain - cybersecurity or networking - rather than a generic business topic.
Common questions
- Discuss the importance of cybersecurity in modern infrastructure
- Discuss how content delivery networks improve web performance
- General current-affairs or technology-trend prompts
Technical Interview(s)
Section titled “Technical Interview(s)”Coding and OOP fundamentals in the first pass - string reversal/replacement, first non-repeating character, operator overloading, smart pointers in C++ - followed by rounds that go genuinely deep on Computer Networks and IT Security, reflecting Akamai’s CDN/edge-security business.
Common questions
- Reverse a string or find the first non-repeating character in it
- Implement operator overloading for a custom class (e.g. complex numbers) in C++
- Explain how a smart pointer works and why you’d use one over a raw pointer
- Walk through the OSI model and where a CDN operates in it
- Explain common web/network security threats and mitigations
Full technical narratives are on the Akamai interview experience page.
Managerial/HR round
Section titled “Managerial/HR round”A closing round that often revisits Linux/OS, Computer Networks, DBMS, DS, and OOP fundamentals at a slightly higher level, followed by a brief HR confirmation call on background and program details - short and informal, usually under 30 minutes.
Common questions
- Explain a Linux command you use often and what it does
- Walk through DNS resolution end to end
- Tell me something about yourself that isn’t on your resume
- Why are you interested in infrastructure/edge computing over a consumer product company?
Sample answer frameworks for each of these are on the Akamai HR interview questions page.
Why networking and security depth matters at Akamai
Section titled “Why networking and security depth matters at Akamai”Unlike most product companies where networking is a light MCQ topic, Akamai builds and operates a large share of the internet’s CDN and edge-security infrastructure - so its technical and managerial rounds treat OSI layers, TCP/IP, DNS, HTTP/HTTPS, and IT security as core subject matter, not a side topic. Candidates who only prep DSA and OOP tend to do fine in the first technical round but get caught out in the networking-heavy later rounds. Budget real study time for CN and basic security concepts alongside coding.
Common technical interview questions and answers
Section titled “Common technical interview questions and answers”Q: How do you find the first non-repeating character in a string?
Make two passes. In the first pass count the occurrences of each character in a hash map, or in a fixed array of size 256 for ASCII. In the second pass walk the original string in order and return the first character whose count is 1, returning a sentinel if none qualifies. That is O(n) time and O(k) space where k is the alphabet size, so O(1) for a fixed alphabet. Iterating the map instead of the string in the second pass is a common mistake, because most hash map implementations do not preserve insertion order and you lose the “first” requirement. A single-pass variant stores the index alongside the count and then takes the minimum index among count-1 entries, which is useful when the string is a stream you cannot rewind.
Q: How does operator overloading work in C++, and when should you use it?
Operator overloading gives operators a meaning for user-defined types by defining functions named operator followed by the symbol. For a Complex class you would write Complex operator+(const Complex& rhs) const as a member, returning a new object by value and marking it const because addition should not mutate either operand. Binary operators that need implicit conversion on the left operand, and stream operators, must be free functions instead; the standard output signature is std::ostream& operator<<(std::ostream& os, const Complex& c), returning the stream so calls can chain. Some operators have required forms: assignment, subscript, function call, and arrow must be members, and assignment must handle self-assignment and return a reference to allow chaining. The judgment part matters most to interviewers: overload only where the operator’s conventional meaning maps naturally, as with arithmetic on complex numbers or matrices, because an operator that surprises the reader is worse than a named method.
Q: What are smart pointers in C++, and how do unique_ptr, shared_ptr, and weak_ptr differ?
Smart pointers are RAII wrappers that own heap memory and free it in their destructor, so the memory is released on every exit path, including exceptions, which raw new and delete cannot guarantee. unique_ptr models exclusive ownership; it is move-only, cannot be copied, and has zero space and time overhead compared with a raw pointer, so it should be your default. shared_ptr models shared ownership using a reference count in a separate control block, freeing the object when the count reaches zero; it costs an extra allocation unless you use make_shared, plus atomic increments and decrements on every copy. weak_ptr observes a shared_ptr without incrementing the strong count and must be converted with lock before use, which is exactly how you break the reference cycle that would otherwise leak when two shared_ptr objects point at each other, for example a parent and child node. Prefer make_unique and make_shared over raw new for exception safety and, in the shared case, a single allocation.
Q: Walk through the OSI model and say where a CDN operates.
The seven layers from the bottom up are Physical, which moves bits over a medium; Data Link, which frames bits and addresses nodes on a local segment using MAC addresses, the layer switches work at; Network, which routes packets between networks using IP addresses, the layer routers work at; Transport, which provides end-to-end delivery through TCP or UDP with ports; Session, which manages dialogue and reconnection; Presentation, which handles encoding, compression, and encryption; and Application, which carries protocols like HTTP, DNS, and SMTP. A CDN operates chiefly at Layer 7, since edge servers terminate HTTP and HTTPS, apply caching rules, rewrite headers, and run application-layer logic like a web application firewall. But it also depends on lower layers: request routing uses DNS at Layer 7 and anycast BGP at Layer 3 to steer users to a nearby point of presence, and TLS termination sits at the presentation boundary. Load balancing is worth distinguishing here, since a Layer 4 balancer only forwards on IP and port while a Layer 7 balancer can route on URL path, cookie, or header.
Q: Walk through DNS resolution end to end.
The browser first checks its own cache, then the OS stub resolver’s cache and the hosts file. On a miss the stub sends a recursive query to a configured resolver, typically the ISP’s or a public one. That resolver checks its cache, and if it has nothing it queries iteratively: first a root server, which returns a referral to the .com TLD nameservers; then a TLD server, which returns a referral to the domain’s authoritative nameservers; then the authoritative server, which returns the actual A or AAAA record. The resolver caches the answer for its TTL and returns it to the client. Queries default to UDP port 53 and fall back to TCP when the response exceeds the size limit or for zone transfers. On a CDN the authoritative answer is usually a CNAME to the CDN’s own domain, and the CDN’s nameservers then answer with the address of a nearby edge point of presence chosen from the resolver’s location and current network conditions, which is exactly how request steering happens. DNSSEC adds signatures so the resolver can verify the chain, and DoH or DoT encrypt the query itself.
Q: How does a CDN actually improve website performance?
Four mechanisms, roughly in order of impact. First, proximity: serving content from an edge point of presence tens of kilometres from the user rather than a single origin thousands of kilometres away cuts round-trip time, which dominates page load because TCP and TLS handshakes each cost round trips. Second, caching: static assets are held at the edge keyed by URL and served on a cache hit without touching the origin, controlled by Cache-Control max-age and validated with ETag or Last-Modified on revalidation. Third, connection optimisation: the edge keeps warm, tuned, persistent connections back to the origin over an optimised network path, terminates TLS locally, and speaks HTTP/2 or HTTP/3 to the client. Fourth, offload and protection: origin bandwidth and CPU drop sharply with a high cache hit ratio, and the distributed edge absorbs volumetric DDoS traffic. For dynamic content that cannot be cached outright, techniques like edge compute, stale-while-revalidate, and micro-caching for a few seconds still remove most origin load.
Q: What are the most common web security threats, and how do you mitigate them?
SQL injection happens when user input is concatenated into a query; the fix is parameterised statements or prepared queries, never string escaping alone, plus least-privilege database accounts. Cross-site scripting injects attacker script into a page; mitigate by context-aware output encoding, a strict Content-Security-Policy, and marking session cookies HttpOnly so script cannot read them. Cross-site request forgery makes a logged-in user’s browser issue a state-changing request; mitigate with per-session anti-CSRF tokens and SameSite cookie attributes. Insecure direct object references, where an endpoint checks that you are logged in but not that you own the record, are fixed by authorisation checks on every object access. Volumetric and application-layer DDoS is absorbed by a distributed edge with rate limiting, and man-in-the-middle attacks are prevented by HTTPS everywhere with HSTS. The defence-in-depth principle matters more than any single control: validate input, encode output, authenticate and authorise separately, and assume any one layer can fail.
Q: Which Linux commands would you use to debug a slow or unreachable server?
Start with the machine itself: top or htop for CPU and memory pressure, and importantly the load average compared against core count; free -h for memory and swap use; df -h and du -sh for a full filesystem, which is a common silent cause of failures; and iostat or vmstat for disk wait. For the process, ps aux plus lsof -p to see its open files and sockets, and strace to see which syscall it is stuck in. For the network, ping for reachability and round-trip time, traceroute or mtr to find where latency or loss appears along the path, dig or nslookup to confirm DNS resolves as expected, ss -tulpn to see which process listens on which port, and curl -v or curl -w with a timing format to break a request down into DNS, connect, TLS, and time-to-first-byte. Then read logs with journalctl -u for the unit and tail -f plus grep or awk on application logs. The habit interviewers look for is narrowing systematically from “is the host healthy” to “is the process healthy” to “is the network path healthy”, rather than guessing.
Frequently asked questions about Akamai interviews
Section titled “Frequently asked questions about Akamai interviews”What is the Akamai interview process for freshers?
Akamai typically runs 3-5 rounds depending on the drive: 1. Aptitude test (some drives) - 30 min, general reasoning. 2. Online Assessment on HackerRank (60-75 minutes) - CS fundamentals MCQs (OS, DBMS, networks, OOPs), an SQL question, and 2 coding problems. 3. Group discussion (some drives) - 5-candidate groups on topics like cybersecurity/networks. 4. Technical interview(s) (45-60 minutes) - DSA/OOP coding plus deep Computer Networks and IT Security questions. 5. Managerial/HR round (20-45 minutes) - Linux/OS/CN/DBMS review plus fit and offer discussion.
What questions are asked in Akamai interviews?
Expect standard DSA problems (string reversal/replacement, first non-repeating character) plus OOP questions like operator overloading and smart pointers in C++. Because Akamai runs one of the world’s largest CDN and edge networks, technical and managerial rounds lean noticeably more on networking and security depth - OSI layers, TCP/IP, DNS, HTTP/HTTPS, CDN concepts, IT security - and Linux/Unix command-line comfort than a typical product-company interview.
Does Akamai have a group discussion round?
On some campus drives, yes - groups of about 5 candidates discuss topics like cybersecurity or networking before the technical interview. Other drives skip straight from the online assessment to technical interviews, so check your specific drive’s format.
How many rounds are there in the Akamai interview?
Most Akamai fresher drives run 3-5 rounds: an optional aptitude test, an online assessment, an optional group discussion, one or two technical interviews, and a closing managerial/HR round. Funnels are steep - one reported drive went from 80 candidates down to 2 final offers.
How should I prepare for Akamai interviews?
Practice medium-level DSA and revise OS/DBMS/OOPs for the assessment, but don’t skip networking fundamentals - OSI layers, TCP/IP, DNS resolution, HTTP/HTTPS, and how a CDN works are genuinely likely to come up given Akamai’s domain. Being comfortable with Linux/Unix basics and IT security concepts, and having a clear point of view on ‘why infrastructure/edge computing over a product company’ for the HR round, also helps.