AI IELTS speaking examiner built with NestJS and Gemini

On October 3, 2026, a developer published a technical write-up on DEV Community about building the speaking module for SmoothMock, an IELTS mock-test platform. The author describes IELTS Speaking as a live interview that runs 11 to 14 minutes: an examiner asks a question, the candidate answers, and the examiner picks the next question based on what the candidate just said. The test has three parts and produces a single band score for the whole session.

According to the post, Speaking was the hardest module to build for SmoothMock. A fixed list of recorded questions would have been easy to implement, but it could not reproduce the follow-up behavior of a real examiner. Instead, the author chose NestJS and Gemini to create a turn-based AI examiner that reads what the candidate has just said and selects the next question from there, rather than playing pre-recorded prompts in a fixed order.

The report covers only the development approach. It does not say whether the module has since shipped, whether any testing data or user feedback exists, or which specific Gemini model version and API details were used. As of this report, the project stands at the point where the author has made public his approach to a turn-based AI examiner built with NestJS and Gemini.