Meeting Transcription / Real-time speech AI
Conversations into useful context.
Live transcripts, speaker identification, and meeting intelligence connected to Slack workflows.
Whisper processes speech while WebSockets carry live transcription updates.
Illustration of the transcription workflow. Transcript examples are illustrative, not recordings of real meetings.
- Whisper
- PEFT
- WebSockets
- FAISS
- Slack API
- GitHub Actions
- AWS CloudFormation
Results & scope
- Reported transcription streaming latency
- <300 ms
- CV project result
- WebSocket-based live transcription in the September 2024–August 2025 project.
Method & conditions
Self-reported in my CV. The precise timing boundary and hardware are not specified. Not a full speech-to-summary latency claim.
- Business-critical service test coverage
- 80%
- CV project result
- Reported unit-test coverage for the meeting platform’s business-critical services.
Method & conditions
Self-reported in my CV. Not a whole-product or end-to-end coverage figure.
How I approached the work
Engineering decisions.
Separate live updates from meeting intelligence
- Problem
- Transcripts need to arrive during the conversation while collaboration and meeting-assistant tasks continue.
- Decision
- Used Whisper and WebSocket streaming within an event-driven platform, with FAISS-backed speaker identification and transcript-based LLM features.
- Result
- Live transcripts supported summaries, action items, searchable conversations, and Slack meeting and translation integrations.
The work.
I developed a serverless meeting platform around Whisper transcription and real-time collaboration. The work spans the speech pipeline, live updates, Slack integrations, and LLM features that turn a transcript into summaries and action items.
My contribution
Live speech pipeline
Streamed transcription updates over WebSockets and fine-tuned Whisper with PEFT for Vietnamese and other underrepresented languages.
Speakers and meaning
Built speaker identification using FAISS, plus transcript-based summaries, action items, and searchable meeting conversations.
Inside the workflow
Developed Slack meeting and translation integrations, event-driven collaboration services, and automated infrastructure delivery.