Featured Work

Scribe

Scribe
pythonpython
dockerdocker
discorddiscord
geminigemini
kuberneteskubernetes

Operational Overview

Distributed transcription pipeline: Discord capture + pyannote diarization + Whisper transcription + Gemini summaries, with GPU worker nodes for heavy compute and tight Discord integration.

Scribe turns Discord calls into usable artifacts: diarized transcripts + structured summaries — using a distributed architecture so heavy compute runs where it should (GPU worker nodes).

Pipeline

  1. Discord bot joins voice channels and captures/uploads jobs.
  2. Server coordinates jobs and storage.
  3. Worker nodes pull pending work and run: pyannote → Whisper → Gemini.
  4. Results are posted back and tightly integrated with the Discord UX.

System Specifications

  • Year2026
  • CategoryAI / Automation

Featured Breakdown

Problem: Always-on transcription is operationally annoying: diarization/transcription need GPU power, but coordination needs to be cheap and reliable.

Approach: Split into Discord ingestion + coordinator server + GPU workers, and implement diarization → transcription → summary.

Impact: Always-on capture with high-quality output without paying the “GPU box running 24/7” tax.