Rammenta.com
A memory layer for your life and work.
A native iOS, watchOS, and macOS app that turns talk into memory: record or import any conversation and instantly get a clean, speaker-labeled transcript with an AI summary, action items, and the moments that matter, then ask questions across everything you've captured. Every model runs on the user's own device, fully offline; their data never leaves their device.
- Full on-device speech-AI stack: designed and built the entire pipeline to run locally in Core ML: transcription (WhisperKit / Apple SpeechAnalyzer), speaker diarization (FluidAudio), and biometric voiceprint identity (ECAPA-TDNN), with no servers, no external APIs, and no per-use cost.
- Private, instant search and insight: on-device vector search and RAG over the user's entire memory layer (local embeddings + sqlite-vec), with summaries, action items, and decision points generated by Apple Foundation Models, zero cloud dependency; works offline.
- The architecture advantage: a custom-built stack where every AI layer runs on the device erases cloud cost and the privacy and compliance risk of sending sensitive content off device.
Open Rammenta →
K9 Operations
The operating system for pet-care facilities.
Replaced the spreadsheets, sticky notes, and manager memory that ran the resorts: automated checklist generation via Gingr APIs, inventory, labor-capacity analysis, and a CRM. Piloted in production. React 18/Vite on Supabase (Postgres/RLS); iOS + web. ~345k LOC, ~1,935 commits, 33 edge functions.
Open K9 Operations →
