Amazon SDE Intern OA Problems 2026
- Posted
- 20 Feb 2026
Summary
I recently completed an Online Assessment for an Amazon SDE Intern position, which featured two distinct algorithmic problems. I provided my attempted solutions and reflected on the challenges faced.
How it went
Hey everyone, I recently appeared for an Amazon SDE Intern OA and wanted to discuss two problems I faced. I solved them partially but would really like to see cleaner or more optimal approaches from the community.Also i need some guidance how to improve in such OAs , i have mentioned mistakes i made here .
Problem 1 — EC2 Instance Allocation Cost
We are given an array representing the number of instances available for each machine type. Customers arrive one by one (M customers total).
For each customer:
- Allocate an instance from the machine type having the maximum available instances.
- The cost incurred by that customer = (maximum instances before allocation) + (minimum non-zero instances currently present).
- After allocation, reduce that machine’s instance count by 1.
- Continue until all customers are processed or instances run out.
We need to return the total cost incurred by all customers.
Example:
Instances = [1, 3, 2, 4]
Customer 1 → max=4, min=1 → cost=5 → array becomes [1,3,2,3]
…and so on.
Curious about:
- Best data structure choice?
- Any mathematical optimization beyond simulation?
I actually got 8/15 test cases for this problem using greedy sorting approach but i think treemap would have been better here .
Problem 2 — Frequent Item Pair from Orders
Given a list of order strings like:
["B03 B06 B08", "B06 B08", "B12 B06"]
Each string contains product IDs bought together.
Task:
Generate all unique unordered pairs of products bought together.
- If an order has exactly 2 items → that pair itself counts.
- If more than 2 items → generate all combinations of size 2.
Count frequency of each pair.
Return the pair with highest frequency.
If multiple pairs have same frequency → return lexicographically smallest pair.
Example:
"B03 B06 B08" → (B03,B06), (B03,B08), (B06,B08)
Interested in:
- Cleaner implementation ideas
- Handling large inputs efficiently
- Any similar known problems/patterns
This problem is where i made silly mistakes tried brute force the constraints was 50 , but because of intense pressure i made a silly string length calculation mistake instea of 7 string length i used 2 for a pair length and could not figure out where i went wrong so 0/15 test cases passed :(
Questions asked (2)
- 01
EC2 Instance Allocation Cost
Data Structures & Algorithms
We are given an array representing the number of instances available for each machine type. Customers arrive one by one (M customers total).
For each customer:
- Allocate an instance from the machine type having the maximum available instances.
- The cost incurred by that customer = (maximum instances before allocation) + (minimum non-zero instances currently present).
- After allocation, reduce that machine’s instance count by 1.
- Continue until all customers are processed or instances run out.
We need to return the total cost incurred by all customers.
Example:
Instances = [1, 3, 2, 4]
Customer 1 → max=4, min=1 → cost=5 → array becomes [1,3,2,3]
…and so on. - 02
Frequent Item Pair from Orders
Data Structures & Algorithms
Given a list of order strings like:
["B03 B06 B08", "B06 B08", "B12 B06"]
Each string contains product IDs bought together.
Task:
Generate all unique unordered pairs of products bought together.
- If an order has exactly 2 items → that pair itself counts.
- If more than 2 items → generate all combinations of size 2.
Count frequency of each pair.
Return the pair with highest frequency.
If multiple pairs have same frequency → return lexicographically smallest pair.
Example:
"B03 B06 B08" → (B03,B06), (B03,B08), (B06,B08)