FOUNDER PROJECT
Unspiral
AI-powered emotional coach that helps you decode anxiety, overwhelm, and avoidance. Solo-built from concept to Google Play.
Onboarding Flow
A warm introduction that explains what Unspiral is and how it helps
Welcome
The hook that resonates
Chat Preview
See the AI coach in action
Journal
Conversations become insights
Patterns
Track emotional patterns over time
How It Works
From emotional overwhelm to clarity in one conversation
Start a conversation
Open the app when you notice anxiety, overwhelm, or avoidance. The AI coach meets you where you are with warm, precise questions.
Decode the signal
The coach classifies what you're experiencing: intuition, anxiety response, intrusive thought, avoidance, or emotional processing.
Get a journal entry
Each conversation is summarized into a structured journal entry with your emotion, threatened need, and a concrete next step.
Signal Classification
Grounded in behavioral psychology — every response is mapped to a type of emotional signal
Intuition
A calm, quiet knowing. Present-moment and rational. The coach helps you honor it.
Anxiety Response
Jumping to worst case without thinking through resolution. "Under-thinking" disguised as overthinking.
Intrusive Thought
Loud, persistent, irrational. The coach labels it and redirects — no analysis needed.
Avoidance
Self-sabotage meeting an unconscious need. The coach finds the need underneath.
Emotional Processing
A legitimate feeling that needs to be felt, not fixed. Grief, sadness, anger as healthy responses.
What I Built
Solo full-stack development from architecture to app store submission
Mobile App
- React Native + Expo with file-based routing
- Glassmorphism design system (lavender + frosted glass)
- Streaming AI chat with real-time responses
- Animated onboarding with 5 screens
- Swipeable calendar for journal navigation
- Drawer navigation with conversation history
AI & Backend
- Claude Sonnet for chat + emotion classification
- Claude Haiku for structured journal generation
- Prompt caching for cost optimization
- Tool-use loop for persistent user memory
- Supabase Edge Functions (Deno runtime)
- Row Level Security for data isolation
Key Challenges Solved
Streaming AI in React Native
Built custom streaming handler for Supabase Edge Functions that delivers real-time token-by-token responses to the mobile client
Consistent emotion classification
Designed a 5-type signal taxonomy with structured prompts, ensuring the AI produces reliable classifications grounded in psychology
Cross-conversation memory
Implemented a tool-use loop where the AI saves relevant facts during chat, then loads them into context for future conversations
Journal generation from chat
Used Claude Haiku with structured JSON extraction to convert freeform conversations into categorized journal entries