calmdownoscar

Problem-driven architecture, technical discipline, and real-world system impact across studio projects.

Mac App Store · iOS Awaiting Review
Wedding Concierge app icon

Wedding Concierge

macOS iOS Vercel Proxy CloudKit Local MLX Evals

Multi-platform conversational AI planning companion engineered with a secure Vercel API proxy wall, CloudKit cross-device sync, and structured spatial planning tools.

01 · Problem

Planning Decision Fatigue

Wedding coordination presents overwhelming decision fatigue across venue search, budget creation, guest lists, and seating arrangements. Raw LLM chatbots fail to structure unstructured advice into actionable event planning timelines.

02 · Architecture & Protocols

Vercel Proxy & CloudKit Sync

Hybrid cloud architecture combining a secure Vercel API proxy wall (protecting API credentials and enforcing rate limits) with CloudKit cross-device state synchronization (search venues on iOS, manage guest lists on Mac). Prototyped with local MLX LLMs and features deep UI planning tiles that transform conversational prompts into concrete schedules.

03 · Impact

Actionable Event Architecture

Delivers enterprise-grade conversational AI planning with real-time venue search, zero user database tracking, and seamless multi-device sync.

Mac App Store · Open Source
Unli Rice app icon

Unli Rice

macOS iOS MCP Protocol Multi-LLM State Open Source

Open-source persistent memory layer for multi-agent LLM systems over Model Context Protocol (MCP) with on-the-go iOS voice-to-vault synchronization.

01 · Problem

Context Drift & Token Waste

Building complex apps with multiple LLM engines (Gemini, Claude, Codex, Antigravity) causes rapid context window ballooning, expensive token consumption, and hallucination loops when starting fresh sessions. Standard note systems become unorganized bloatware during multi-agent handoffs.

02 · Architecture & Protocols

MCP Protocol & Shared Vault

Engineered an open-source persistent memory layer using Model Context Protocol (MCP) and file-system state synchronization. Connects disparate LLMs to a single ground-truth vault featuring explicit guardrails, evaluation rules, and non-destructive folder governance. Includes an iOS voice-capture app that syncs raw audio notes to macOS for instant agent retrieval.

03 · Impact

Zero Re-explanation Friction

Eliminates context re-explanation across multi-agent chat sessions, enforcing strict architectural consistency and zero-friction model swapping without loss of project memory.

Being Reworked
ClearSpace app icon

ClearSpace

iOS PhotosKit Swift / SwiftUI 100% Local ML Zero Telemetry

Native Swift & PhotosKit rebuild for on-device iPhone photo library analysis, high-throughput media indexing, and privacy-first storage recovery.

01 · Problem

iCloud Bloat & Large Libraries

Managing massive personal photo libraries (40,000+ items per user) leads to escalating iCloud subscription costs and device storage depletion, while third-party utilities suffer from slow cloud processing or privacy risks.

02 · Architecture & Protocols

PhotosKit On-Device Indexing

Native Swift/SwiftUI rebuild utilizing Apple's PhotosKit framework for high-throughput on-device media analysis. Ranks heavy video containers and burst assets locally with zero third-party trackers or external cloud dependencies.

03 · Impact

Privacy-First Fast Scan

Delivers liquid-smooth local scanning of massive media libraries with complete privacy preservation and bulk removal to iOS Recently Deleted for 30-day non-destructive recovery safety.

In Development
Unli Disk app icon

Unli Disk

macOS AppKit / SwiftUI PhotosKit Core Dev Cache Engine

Native macOS storage map & developer cache analyzer. Intelligently identifies orphaned Git repos, build caches, and heavy media assets.

01 · Problem

Developer Cache & Repo Bloat

Developer Mac storage rapidly vanishes under unlinked Git repositories, build artifacts, and hidden caches. Traditional visualizers obscure developer file context, while automated cleaners pose catastrophic data loss risks for uncommitted repos.

02 · Architecture & Protocols

Tree Mapping & Dev Rules

High-performance macOS disk scanner built with AppKit/SwiftUI that generates transparent disk tree maps and detects orphan repositories. Integrates the shared Packages/UnliDiskPhotos engine for local photo auditing and enforces strict safety by routing files exclusively to macOS Trash.

03 · Impact

Safe Disk Space Reclamation

Reclaims gigabytes of developer disk bloat safely without blind automated deletions or project corruption risks.

Draft Stage
A

Architecturally

iPadOS 3D Geometry Engine Generative AI Rendering Spatial Data

iPad sketch-to-3D floorplan engine. Transforms freehand touch/stylus strokes into parametric geometry and AI visual renders for spatial planning.

01 · Problem

Rigid CAD Software Limits

Traditional CAD tools demand steep learning curves and rigid inputs, hindering rapid architectural exploration for complex custom floorplan requirements (e.g. child safety zones, pet areas, vehicle storage).

02 · Architecture & Protocols

Parametric Sketch Interpreter

Sketch-first iPad design system that interprets freehand stylus strokes into structured parametric floorplan data, automatically generating real-time 3D building models and AI visual renders from a unified spatial model.

03 · Impact

Rapid Spatial Prototyping

Accelerates early-stage spatial ideation from rough napkin sketches to interactive 3D geometry in seconds.

From voice capture to multi-agent pre-mortems, backend architecture, security audits, and post-mortems — with automated return verification loops at every gate.

Unli Rice Voice Capture (iOS) Unli Rice Vault (Notes) Gemini Spark (Feasibility & Market) Claude (Pre-Mortem & Evals) Unli Rice / MCP Vault Sync Codex (Backend Architecture) Antigravity (UI/UX Frontend) Figma / MCP Media Claude (Security Audit & Encryption) Codex (Post-Mortem & Regression) Vercel Proxy & App Store Release Unli Disk (Frees storage by deleting simulators & dead end repos) Verification Return Hardening Loop Regression Check

A multi-agent development methodology: voice ideation → feasibility benchmarks → pre-mortem analysis → backend architecture → UI/UX mockups → security audits & post-mortems.

01 · Voice Capture & Vault Ingestion

Ideas are spoken directly into the Unli Rice iOS capture app and transcribed to markdown text. Notes automatically sync to the local Unli Rice vault as ground-truth memory accessible by connected MCP agents.

02 · Feasibility & Market Research

Gemini Spark ingests the raw vault text, analyzes technical feasibility, and benchmarks existing market solutions to evaluate whether the concept solves a genuine operational gap.

03 · Pre-Mortem & Evals Injection

Claude performs a pre-mortem failure analysis, adding context rules, guardrails, and evals. Once approved, Claude generates a handoff prompt for Codex to build backend schemas and core logic, returning for verification.

04 · UI/UX Assembly & Security Audit

Antigravity builds UI/UX mockups and frontend components (with Figma/Higgsfield over MCP). Near completion, Claude runs a security audit for exploit vectors and encryption, followed by a Codex post-mortem before Vercel proxy & App Store release.