Kotak | SDE 2 | Reject

Experience: ~4 Years

Round 1 – DSA + Discussion

We started with a general discussion about my job, the most challenging work I’ve done, and related experience.

Problem Statement :

Given a tree T with N nodes, where each node i has Ci coins attached to it:

You need to choose a subset of nodes such that:

No two chosen nodes are adjacent (i.e., directly connected by an edge)

The sum of coins from the chosen nodes is maximized

Input

Number of nodes: 4

Edges:

0 - 1 0 - 2 2 - 3

Coins:

[2, 5, 3, 1]

Output 8

Explanation: We can choose nodes 1 and 2 with coins 5 and 3.

My Approach

Used DFS + Dynamic Programming

For each node, computed:

Max sum if the node is included

Max sum if the node is excluded

Time Complexity: O(N) Space Complexity: O(N)

The interviewer seemed satisfied with the solution.

The round ended with a few questions on: HLD and Java Internals

---

Round 2 – Concurrent Systems (Java) Problem Statement :

You are asked to design a Concurrent Log Processor in Java.

Log entries come as strings in the format:

userId,eventType,timestamp

Example Logs u1,LOGIN,2026-01-01T11:15:10 u2,LOGOUT,2026-04-12T11:16:00 u1,LOGIN,2026-04-11T12:17:10

Multiple threads will push logs concurrently.

Requirements

Implement a class LogProcessor with:

public void acceptLog(String logLine); // Called by multiple threads

public Map<String, Integer> getUserEventCounts(String userId);

acceptLog must be thread-safe

getUserEventCounts("u1") should return something like:

{ "LOGIN": 2, "LOGOUT": 1 }

Solution should be thread-safe and reasonably efficient

My Solution :

Used ConcurrentHashMap

Outer map: userId → eventMap

Inner map: eventType → count

Interviewer was okay-ish with this approach

Follow-up Questions :

Interviewer asked how I would implement ConcurrentHashMap internally

I explained basic ideas like:

• Fine-grained locking

• Locking on writes

• Concurrent reads

However, I lacked deep knowledge of its internal implementation

I also suggested a scheduler-based batching approach

• Accept logs into a queue

• Periodically update aggregates

This introduces slight latency in reads

The interviewer did not seem fully satisfied with these answers.

Final Outcome

The interview ended politely. I received a rejection email after a couple of days.

Request for Feedback :

It would be great if someone could suggest, a better design for the concurrent log processor.

Hope this helps others preparing for similar interviews.