Build a Voice AI Screener That Actually Understands Candidates

Most HR teams first look at tools like MyInterview or HireVue. These platforms work for structured video interviews, but their voice analysis is often keyword-based. They can flag a candidate who says "teamwork" but cannot distinguish between someone who gives a genuine example and someone who just drops the buzzword. They also lock you into rigid question paths and expensive per-seat contracts.

Some teams try to build a simpler version with traditional Interactive Voice Response (IVR) systems. This approach fails because IVRs are just phone trees. They can ask "Press 1 for yes, 2 for no" to confirm a certification, but they cannot handle open-ended questions like, "Tell me about a time you solved a problem on the fly." This misses any chance to assess critical soft skills.

A regional logistics company hiring 50 warehouse staff per quarter faced this exact issue. Their two-person HR team was overwhelmed. The off-the-shelf tools filtered out good candidates who didn't use specific keywords, and the IVR system couldn't tell them anything about a candidate's personality or problem-solving skills, leading to many wasted second-round interviews.

An engagement with Syntora would begin with a discovery phase to understand your specific hiring workflows, job requirements, and desired integration points with your Applicant Tracking System. This informs the custom scoring rubric and prompt engineering for the AI.

The technical approach typically starts by provisioning a dedicated phone number using Twilio. When a candidate calls, the audio would be streamed to an AWS S3 bucket. An AWS Lambda function would then trigger on the file upload and transcribe the conversation efficiently using OpenAI's Whisper API, providing a clean, time-stamped text transcript.

This transcript would then be processed by the Claude 3 Sonnet API. Syntora designs a multi-shot prompt that incorporates the specific job description, your custom scoring rubric with defined key traits (e.g., communication clarity, problem-solving skills), and examples of strong and weak answers. Claude then scores each trait and provides a written justification, referencing specific parts of the conversation.

The output is a structured JSON object containing scores, justifications, and the full transcript. Syntora would integrate this data directly into your existing ATS, such as Greenhouse or Lever, using their native APIs. The aim is for recruiters to access a summary score and transcript notes on the candidate's profile shortly after the call concludes.

The system architecture would be serverless, built with Python on AWS Lambda and connected via Amazon SQS queues, eliminating the need for server management. For systems processing up to 500 candidates monthly, projected AWS compute costs are typically low. A typical engagement for a system of this complexity, from initial discovery to deployment of a working prototype, usually spans 4-6 weeks, with ongoing iteration and support available.