The Brief
The goal of this project was to build voice agent that automate pre-screening. Quantifiable questions — visa, experience, salary, start date, etc — parsed and surfaced to the recruiter. Open-ended behavioral questions summarized into a candidate brief. Recruiters get structured signal without sitting through hours of calls.
Can be integrated natively into any ATS, so recruiters rarely have to leave their ATS of choice.
What I Built
Owned this start to finish — architecture, infrastructure, delivery, scope.
Simple scope. We kept it that way.
Serverless on AWS Lambda, Terraform for infrastructure, ElevenLabs for voice agents, Auth0 on the auth layer.
Results
Monthly running cost for the MVP: cheaper than a lunch.
Shipped within time and budget. The hardest part wasn’t the tech — it was holding the line on scope when it started to creep. The result is a lean, secure, extensible product that does exactly what it was supposed to, nothing more.
Still running. Still cheap. Ready to scale.
Tech stack
Frontend: React, Vite, Tailwind, MobX
Backend: Node.js, Serverless Framework, Drizzle ORM, PostgreSQL, Zod, AWS SES
Infra: AWS Lambda, AWS S3, CloudFront, Terraform, Terragrunt, Docker, GitHub Actiona
AI: ElevenLabs, LangChain, OpenAI
Others: Auth0, TypeScript, Turborepo, pnpm