CalmScribe: On-Device AI Journaling Application

CalmScribe is a privacy-first mobile note-taking and journaling application that processes natural language entirely offline. By leveraging edge computing and local Large Language Model (LLM) inference, the application provides users with immediate, intelligent emotional analysis of their journal entries without ever sending personal data over a network.

Technical Architecture

Key Engineering Milestones

1. Local LLM Deployment

Successfully navigated complex licensing and network gating to host a 1.5-billion-parameter neural network directly within the mobile application sandbox.

2. Zero-Network Processing

Implemented an ephemeral chat session architecture that allocates device RAM on demand, providing deep emotional insights and thematic summaries with zero API costs and absolute data privacy.

3. Native Build Resolution

Engineered robust deployment pipelines to resolve advanced Swift Package Manager (SPM) caching conflicts and Xcode minimum-target deployment mismatches, ensuring seamless compilation across both Apple Silicon and iOS Simulators.

System Philosophy

At its core, this project bridges the gap between psychological wellness and bleeding-edge mobile engineering. The architecture was built with a focus on resilient, offline-first design—resulting in a robust application capable of interactive self-correction and safe workflow execution entirely at the edge.