LastPush - AI Based Voice Interview Platform for Candidate Pre Screening

The recruitment pre-screening process is often characterized by inefficiencies, human bias, and significant operational costs. Existing digital solutions, such as form-based or video interview platforms, frequently fail to offer a natural, conversational experience, leading to poor candidate engagement and inconsistent evaluations. To address these limitations, this paper presents LastPush, an AI-based voice interview platform that leverages speech-to-text (STT), large language models (LLMs), and text-to speech (TTS) technologies to conduct real-time, guided voice interviews. The platform operates across three primary phases – Setup, Capture, and Analysis. Orchestrated through a serverless architecture integrating Next.js, Supabase, and the Gemini API. During the capture phase, the Vapi.ai SDK manages the conversational flow, while the MediaRecorder and PageVisibility APIs enable browser-based proctoring to ensure authenticity. Captured data forms a Synchronized Interview Log, which is evaluated by Gemini using an LLM-as-a-Judge methodology for unbiased scoring. This architecture demonstrates a scalable, bias-mitigated, and cost-efficient approach to recruitment automation.