What candidates say the round looked like, summarised in our words, each linked to where it came from. Razorpay hasn’t confirmed these details unless the source says “Official page”.
read the files, write code, run tests (at least half of 1 reports)
Length
about 90 min (median of 1 reports)
The reports
Onsite · June 2026
Mid-level · Own machine · 90 min
SDE-2 agentic coding round: 90 minutes, you own the design and implement it with your own AI coding agent, bringing your own keys and tooling. The candidate used Claude Code with custom subagents and a personal conventions file. Graded: the low-level design, how precisely you prompt the agent, process discipline (small commits, reviewing every diff, testing as you go) and whether the final code works. Get fluent with your setup beforehand, since the clock starts when the problem is shared. General tooling is fine, but anything pre-built for specific problem types gets flagged.
AI: Your own AI tools that could read the files, write code, run tests · Task: Build from a spec · Source: LeetCode Discuss post