This is micro1's internal tool that we built to save our internal team's time during interviews. Candidates write their top 3 tech stacks and their expertise per tech stack. We then use that to generate custom questions using gpt-4. We ask the candidate to use their voice to answer questions and convert that back to text using Whisper API. We then send that text back to OpenAI for an assessment. We also have a proctoring layer and AI classifier layered on top of the software.
Our main challenge was to minimize the probability of cheating during the candidate's gpt-vetting. To address this, we implemented multiple layers of security measures, such as acquiring software to detect mouse movement and eye movement, utilizing an AI classifier, and a timer for each question that starts right when the candidate sees the question.
This project is complete but it is pending to go live. We will update the status once it's live.