Last year I was in a crucial client meeting discussing a potential $2M partnership. I was so focused on appearing engaged that I missed the key technical requirements they mentioned.
When I followed up with generic questions, they went with a competitor who "clearly understood their needs better."
That weekend, I was furious. I decided to build something to never let this happen again.
*Technical stack (built in 72 hours):*
- Swift + AVAudioEngine for audio capture
- OpenAI Whisper API for transcription (99%+ accuracy even in noisy rooms)
- ChatGPT API for intelligent note structuring and action item extraction
- Core Data for local storage (privacy-first approach)
- Background processing to handle long meetings without draining battery
The AI workflow:
1. Records audio locally in chunks
2. Streams audio to Whisper API for transcription
3. Raw transcripts fed to ChatGPT every 30 seconds to:
- Identify key decisions and action items
- Tag speakers and topics
- Generate concise summaries
- Flag follow-up questions
Interesting technical challenges:
- iOS kills background processes aggressively, solved with silent audio trick
- Chunking audio without cutting words mid-sentence (used silence detection)
- Handling poor audio quality in conference rooms (noise filtering + gain adjustment)
- API rate limiting during long meetings (intelligent batching)
Day 1: Core recording + Whisper integration
Day 2: ChatGPT processing + note structuring
Day 3: UI polish + App Store submission
The irony? That same client reached out last month after seeing how prepared I am in recent meetings. We're now discussing an even bigger partnership.
Would love your technical feedback!
When I followed up with generic questions, they went with a competitor who "clearly understood their needs better."
That weekend, I was furious. I decided to build something to never let this happen again.
*Technical stack (built in 72 hours):* - Swift + AVAudioEngine for audio capture - OpenAI Whisper API for transcription (99%+ accuracy even in noisy rooms) - ChatGPT API for intelligent note structuring and action item extraction - Core Data for local storage (privacy-first approach) - Background processing to handle long meetings without draining battery
The AI workflow: 1. Records audio locally in chunks 2. Streams audio to Whisper API for transcription 3. Raw transcripts fed to ChatGPT every 30 seconds to: - Identify key decisions and action items - Tag speakers and topics - Generate concise summaries - Flag follow-up questions
Interesting technical challenges: - iOS kills background processes aggressively, solved with silent audio trick - Chunking audio without cutting words mid-sentence (used silence detection) - Handling poor audio quality in conference rooms (noise filtering + gain adjustment) - API rate limiting during long meetings (intelligent batching)
Day 1: Core recording + Whisper integration Day 2: ChatGPT processing + note structuring Day 3: UI polish + App Store submission
The irony? That same client reached out last month after seeing how prepared I am in recent meetings. We're now discussing an even bigger partnership.
Would love your technical feedback!