Oracle USA PMTS Offer
Summary
I received an offer for the Principal MTS role at Oracle USA after a series of phone screen and onsite interviews, which I ultimately passed on due to competing staff-level offers.
Full Experience
I interviewed for the Principal MTS (PMTS) level at Oracle USA. The process involved a phone screen followed by a comprehensive onsite loop.
Phone Screen (60 mins)
- I faced 2 behavioral questions.
- There was 1 standard coding question on binary trees, which I completed quickly in about 5 minutes.
- The main focus was a system design problem centered on a metrics and logging system. This round lasted almost 35 minutes and involved extensive discussions on Kafka, Elastic Search DB, various design choices, and tradeoffs.
Onsite Interview Rounds
My onsite interviews included:
- System Design: This round began with 2 behavioral questions. The core technical part was to design a job scheduler system. Having studied this topic previously, the conversation flowed smoothly. The interviewer delved deep into indexes, specifically primary and sort keys of DynamoDB, Global Secondary Indexes (GSI), and the rationale behind using NoSQL for such a system.
- Hiring Manager: This 60-minute session was split into 30 minutes of behavioral questions and 30 minutes dedicated to system design. The design question was closely related to the hiring manager's team's work. They probed extensively into all my design choices, requiring me to propose alternatives and compare tradeoffs, and also discuss the design's flexibility in adapting to future requirement changes.
- Coding Round 1: After 2 behavioral questions, I received a big question prompt on Hackerrank. It involved flights data with dependencies (e.g., flight1 → flight2). The task was to return a sorted list of all delayed flight IDs, given an initial list of delayed flights. I recognized it as a simple DFS application, though I initially considered topological sort.
- Bar Raiser (Bar Tender): This 60-minute session was an in-depth conversation covering my background, most challenging project, design choices, my role and impact, collaboration with other teams, conflict resolution, situations where things went south, and how I mentored junior/mid-senior engineers. It was a lot of talking.
- Coding Round 2: Similar to the first coding round, this also started with 2 behavioral questions. The problem statement was explained verbally, without a prompt on Hackerrank, so I took notes. The question boiled down to finding the maximum length of a subarray with all unique IDs in a given list. A follow-up asked what if the subarray could contain at most 'k' distinct unique IDs. I knew these types of problems are common on LeetCode.
The recruiter contacted me within two days with positive feedback and an offer. However, I ultimately decided to pass on the offer as I had competing staff-level offers from other companies.
Overall, I found the conversations pleasant and the interviewers to be good.
Interview Questions (4)
Design a system for metrics collection and logging. Discuss core components, design choices, and tradeoffs. Specific areas of discussion included Kafka and ElasticSearch DB.
Design a job scheduler system. The discussion delved into details such as indexes, primary and sort keys for DynamoDB, Global Secondary Indexes (GSI), and the rationale for using a NoSQL database in this context.
Given flights data with dependencies (e.g., flight1 → flight2), and a list of flight IDs that are delayed, return a sorted list of all flight IDs that will be delayed due to these initial delays, considering the dependencies.
Given a list of IDs, find the maximum length of a subarray that contains all unique IDs. As a follow-up, consider a variant where the subarray can contain at most 'k' distinct unique IDs.