I applied through a recruiter. The process took 2 weeks.

The process started with Recruiter A reaching out to me on LinkedIn, requesting my resume, and sharing the job description. Two days later I received a scheduling link for a screening call with Recruiter B.

The recruiter screening felt unusual. There was no introduction and we jumped straight into discussing my projects, followed by behavioral questions and broad technical questions on Java, databases, isolation levels, and CQRS. One example was, "You have a database with 100 million users. How would you optimize it?" When I asked about the actual bottleneck or performance problem, I was told to "just tell."

The next day, Recruiter B scheduled another call purely to explain the upcoming interview process, after which Recruiter C sent the scheduling link for the first technical interview. Having 3 different recruiters involved before even reaching the first technical round made the process feel fragmented.

The first technical interview consisted of roughly 35 minutes of coding and 20 minutes of high level discussion. AI tools were explicitly allowed. Since this was my first AI-assisted coding interview, I asked the interviewer what he expected me to write myself vs what should be generated by AI. He said it was entirely up to me and had no preferred workflow.

I proposed a spec-driven approach using an AI coding agent with an open-source model, with implementation and verification control. He suggested using TDD with Spec, which I agreed to. After clarifying the requirements, and telling him the exact design and data structures, he said let's not write plan.md, just put the question and your thoughts in the coding agent to generate specs. And he said ask the coding agent to implement phase 1.

The interviewer then asked me to stop using AI and continue manually. I reviewed the generated code; one function was left and he said implement that. I took 5-10 minutes as that was one of the main functionality functions. He then said we are running low on time, ask your coding agent to finish all remaining phases and discuss how I would design the database if the solution were persistent rather than in memory. I told all with relations.

2 days later I received a rejection stating that I was "not up to speed." I fully accept interview outcomes, but I found the evaluation criteria confusing. In an interview where AI usage is explicitly encouraged, expectations around what candidates should generate vs implement manually should be defined by the interviewer rather than left entirely to the candidate. It also felt contradictory to encourage AI usage and then question that AI generated parts of the solution.

With AI increasingly becoming part of day-to-day engineering, I believe interviews should evaluate:

• Problem solving
• Architecture
• Tradeoff analysis
• The ability to effectively guide and verify AI-generated code

This is much better than focusing primarily on typing speed after AI assistance has been introduced. Overall, the process left me feeling that the AI interview format was still evolving and that clearer expectations and a more streamlined recruiting experience would significantly improve the candidate experience.

Code a Parking lot in Java/Kotlin