RecapRelay
Automation: schedule meeting on Google Calendar; bot auto-joins via Vexa AI; transcribes and takes notes; sends summary to stakeholders. Streamlined meeting documentation.
TODO: Add permission-approved, sanitized product screenshots. Concept art is not shown as product evidence.
Technical concept illustration. The case study below describes the project implementation.
Overview
RecapRelay is an end-to-end meeting automation pipeline: schedule a meeting on Google Calendar, and the bot auto-joins via Vexa AI, transcribes the conversation, takes notes, and sends a summary to stakeholders — with zero manual note-taking.
Why This Was Hard
Each step in the chain (calendar trigger, meeting join, transcription, summarization, distribution) is a separate integration with its own failure modes — the real engineering challenge was making the full pipeline run reliably end-to-end, not just getting each piece working in isolation.
What I Built
- Google Calendar integration — reads scheduled events and meeting details automatically
- Automated meeting join via Vexa AI the moment a meeting starts
- Full transcription of the meeting audio
- AI-driven note extraction — action items, decisions, and key points pulled from the transcript
- Automated summary distribution to stakeholders in a formatted, readable format
- Chained pipeline — calendar trigger → join → transcribe → summarize → send, running without manual intervention
Tech Stack
- Calendar: Google Calendar API
- Meeting join: Vexa AI
- Transcription: Whisper / Google Speech-to-Text
- Summarization: LLM-based (OpenAI/Claude)
- Notification: Email / Slack delivery
Key Decisions
- Chained the pipeline (calendar trigger → join → transcribe → summarize → send) as discrete, independently retryable steps rather than one monolithic job, since each integration point had its own failure mode and needed to fail without breaking the whole run
- Used an LLM for summarization rather than a purely extractive method, since stakeholders needed readable action items and decisions, not a list of transcript excerpts
- Delivered summaries automatically on meeting end rather than requiring a manual trigger, since the entire point was removing manual note-taking from the workflow
Outcome / Impact
Meeting notes and summaries now arrive automatically the moment a meeting ends — no manual note-taking, and stakeholders get an actionable summary without lifting a finger.
Related work
DigestPress
Automated news scraping and summarization tool — pulls articles from multiple sources and generates concise daily digests.
CaseDraft
Internal tool that accepts Figma images or file paths and uses AI to auto-generate test cases and use cases. Python with Selenium for extraction. Output formatted for direct copy-paste into Word.
GuestGallery
Python app using face recognition to allocate 2000+ wedding photos to guests. Finds each person across photos and surfaces “here are your pictures.” Processes large image sets efficiently.
Interested in similar work?
Review my services and how we can define a useful first milestone.
Discuss similar work