Passed Amazon SDE New Grad
Anonymous User
2007
AmazonCareerInterview

Got the L4 New Grad SDE Offer at Amazon – Here's How I Prepared

I recently got an offer for a new grad SDE (L4) position at Amazon, and I wanted to share my journey — from knowing nothing about DSA to cracking the interviews. Hopefully, this helps someone who's starting from scratch too.

Phase 1: Learning the Fundamentals (February)

  • In February, I had no clue about data structures and algorithms.

  • To build a strong foundation, I completed:

    • Stanford’s Algorithm Specialization (Courses 1, 2, and 3; 4 was not necessary):
      Coursera Link

Pros:

  • Great for understanding the theory behind common algorithms.

Cons:

  • Possibly overkill for interviews, but I preferred overpreparing rather than missing key concepts.

Phase 2: Problem Solving (April)

  • Once I had the theory down, I started solving LeetCode problems.
  • I often used AI to help understand solutions when I got stuck, but never just copy-pasted answers.
  • I always made sure I understood the approach.

My path:

  • Started with the LeetCode 75 Study Plan
  • Then moved on to NeetCode 150, solving around 70 problems

Note: NeetCode is one of the best resources for DSA interview prep. Use it strategically.

Phase 3: Online Assessment + Work Simulation (Mid-May)

  • Got an email saying I had 5 days to complete the OA.

Problem 1:

  • Count the number of palindromes in a string (or something similar).
  • My solution didn’t pass all test cases — not because it was wrong, but because it was too slow.

Problem 2:

  • Required a greedy + heap approach.
  • I passed all the test cases.

Work Simulation:

  • Was supposed to be open for 5 days. After one day, I got another email saying the next day was the last.
  • Since it was Saturday and support wasn’t available, I completed it immediately.

Phase 4: First Technical Interview (30 Minutes)

Questions:

  1. Anagram Checker

    • Determine if two strings are anagrams.
    • The interviewer asked me not to use Python’s built-in functions.
    • Still a fairly easy problem.

  2. Stream of Words

    • For each incoming word, return the last seen anagram (if any), or the word itself.
    • I used the same logic from the previous problem and a hash map with sorted keys.

I passed and was invited to the final round: three back-to-back 1-hour interviews.

Phase 5: Final Interviews (3 x 1hr on the Same Day)

Behavioral Preparation (Leadership Principles)

  • Wrote five STAR-format stories that covered most of Amazon’s Leadership Principles.
  • Practiced behavioral answers using questions generated by ChatGPT.
  • Rehearsed with my girlfriend.

Technical Interviews

Interview 1:

  • Deepest Level in a Tree

    • Given a tree (not necessarily binary), return its maximum depth.
    • I used a BFS approach.

  • Lowest Common Ancestor

    • Find the LCA of two nodes in a tree where each node has a pointer to its parent.
    • I solved this by recursively propagating upward.

Interview 2:

  • This round was system design/database oriented, which caught me off guard.

Question:

  • Design a system to track how many people are in the office at any given time.

Follow-ups:

  • Design queries to return the number of people at a specific timestamp.
  • Find the maximum number of people during a time interval.

I didn’t perform well here due to no experience in OOD or DB design. The interviewer wasn’t very kind and laughed a bit when I got stuck. I stayed focused and moved on.

Interview 3:

  • Validate Alexa commands based on rules such as:

    • The first word must be “Alexa”
    • No repeated words back-to-back
    • Other constraints

Initial Approach:

  • Hardcoded all checks using logical conditions.

Refactored Approach:

  • Created an abstract Rule class
  • Defined each rule as a subclass
  • Stored rules in a set and validated them using a loop
    (more scalable and cleaner)

Final Thoughts

  • You don’t need to solve all 150 NeetCode problems.
    Understanding patterns and building intuition is far more important.

  • Use AI to learn, not to cheat.
    Your understanding matters more than the number of problems you “complete.”

  • Some interviewers emphasize how your algorithm works, not just correctness.
    For example, in the BFS problem, I was asked:

    • Why BFS uses a queue
    • The advantages and disadvantages of BFS and DFS
    • When I would use each approach
Comments (9)