Phone Tester
Voice Quality Analysis Platform — Built and deployed by Bruhn Freeman
Project Overview
Phone Tester is a proprietary voice quality verification platform built to answer a deceptively simple question: does the caller actually sound clear on the other end? Remote workers, call center agents, and customer-facing operators rely on their phone and internet setups without ever having an objective way to test them. Phone Tester fills that gap.
The system places a real Twilio voice call to the user, prompts them to speak a set of randomly generated NATO phonetic code words, transcribes the result using speech-to-text, and scores the session. The entire test takes under two minutes and produces a clear pass or fail — with no subjectivity involved.
The Challenge
Organizations deploying remote workers or phone-based customer service teams had no reliable, self-serve way to validate audio quality. Complaints like "I sound muffled" or "my calls keep dropping" were anecdotal and hard to act on. Third-party services that offered any form of voice testing were either cost-prohibitive, generic, or produced results that couldn't be tied back to a specific user's setup.
- No self-service diagnostics existed that could test actual voice transmission quality end-to-end
- IT teams were reactive — issues were only flagged after a client or customer complained
- Existing tools tested bandwidth but not how the human voice was actually received by the other party
- Results needed to be tied to identities — knowing a test passed or failed meant nothing without knowing who ran it
The Solution
We designed and built Phone Tester from scratch as a full-stack web application with Twilio at its core. The platform combines email-verified user registration, a structured voice call workflow, real-time speech recognition, and a persistent results dashboard.
- User Registration: Email + phone verification flow with bcrypt-secured credentials and geo-tracked login history
- Test Initiation: Five random NATO phonetic code words are generated and emailed to the user before the call begins
- Live Voice Call: Twilio places an outbound call to the user's registered number and delivers an IVR prompt
- Speech Recognition: The caller's spoken words are transcribed and matched using fuzzy logic (±1 character tolerance per word)
- Scoring & History: Results are stored per user with timestamps, attempt counts, and status flags
- Suspicious Activity Detection: Geolocation data flags login patterns that deviate significantly from a user's established location history
Technical Architecture
The platform runs as a containerized Node.js application backed by MongoDB, deployed on AWS Lightsail via Docker Compose. Twilio handles all voice and SMS infrastructure.
Key Results
End-to-end test completion in under 2 minutes from registration to pass/fail result — no manual IT intervention required.
- Speech recognition accuracy of 97%+ on standard English phonetic words across tested network conditions
- Replaced a $400+/month third-party voice monitoring service with a self-hosted, fully owned solution
- Geolocation anomaly detection flagged 100% of simulated suspicious logins during internal QA testing
- Supports concurrent test sessions without resource contention — stateless design scales horizontally
- Full SMS compliance built in: opt-in at signup, STOP/HELP keywords, carrier-compliant message cadence
Interested in a Similar System?
Whether you need voice quality monitoring, an identity verification workflow, or a custom communications platform built around Twilio, we engineer it from the ground up — no off-the-shelf compromises.