Building a 20-Hour DBMS Interview Prep System Using LLMs

The author used large language models (LLMs) to create a structured DBMS interview preparation system, including curating questions, finding the minimal set of resources, and using the LLM as a mock interviewer.

Details

The author ran an experiment using LLMs to build a comprehensive DBMS interview preparation system. They first curated a list of 100 frequently asked DBMS interview questions, spanning topics from SQL fundamentals to advanced concepts like transactions, indexing, and replication. Then, they asked the LLM to identify the minimal set of resources that would cover all the questions, which turned out to be InterviewBit articles, selected chapters from the DDIA book, and LeetCode's Top SQL 50 problems. The total estimated preparation time was around 20 hours. The key innovation was using the LLM as a mock interviewer, where it would ask the questions one by one and provide feedback on the quality and depth of the responses, forcing the author to articulate the concepts clearly. The author wants to get feedback on whether this structured approach using LLMs actually translates well in real interviews or if it comes across as rehearsed, and if there are any blind spots in using an LLM for both curriculum design and mock interviews.