Work

Voice ML microservice

Steve's Computers

I built an AI service that recognizes which technician is speaking on a repair shop's shared phone line, transcribes the call, and reports how confident the match is.

Steve's Computer Repair call-processing view with a voice waveform, technician match and transcript.
1day to deliveryquoted four to five
Client
Repair shop
Role
Built
Timeline
2026 · Delivered
Stack
Python, Resemblyzer, faster-whisper, Docker, Coolify

Delivered with a five-star review, and I've worked with the shop again since.

Four technicians share one phone line. The shop wanted transcripts tagged by who was actually speaking, without keeping the audio.

  1. Built AI voice recognition that identifies enrolled technicians on the shop's shared phone line and reports how confident the match is.
  2. Transcribed each call with a speech-to-text model, handled the phone system's audio format, and added a way to improve matching with new examples.
  3. Deleted recordings after creating the voice representations used for matching.
  4. Published a reusable version of the code.

What the project delivered

90%confidence for one recorded speaker matchThe system's confidence in that match, not an overall accuracy rate.

Say hello.

Tell me what you'd like to build or improve.

Quick navigation

Find a page.

Use the arrow keys to choose a result, Enter to open it, or Escape to close.