Inspiration##
I’ve often seen friends and myself stuck in indecision: which job to take, which project to launch, what to prioritize. Good coaching changes everything, but it’s expensive, hard to find, and time-consuming. I wanted to build something instant, affordable, and always available – an AI coach that cuts through the noise and gives real clarity fast.
ClarityCoach is an AI-powered coaching application designed to help people move from confusion to structured action.
Unlike most AI chat apps that begin with a blank input field, ClarityCoach starts with intention.
Before the conversation begins, users answer two questions:
- What are you stuck on?
- What is your priority?
This simple framing dramatically improves the quality, relevance, and structure of the AI’s responses.
The goal is not to generate answers. The goal is to generate clarity.
The Problem
Many people feel overwhelmed — too many ideas, too many goals, too much information.
AI can generate content endlessly. But without structure, it often increases noise instead of reducing it.
I wanted to build an AI experience that behaves more like a structured coach than a chatbot.
The Solution
ClarityCoach enforces intentional context before any AI interaction.
By defining the user's blockage and priority upfront, the AI can:
- Provide focused guidance
- Offer structured action steps
- Encourage reflection instead of generic advice
- Keep the conversation goal-oriented
This transforms AI from reactive answering into proactive coaching.
Technical Architecture
ClarityCoach was built with a scalable and production-ready architecture:
- SwiftUI for a clean and minimal iOS interface
- Python (FastAPI) backend for secure AI communication
- Gemini API for structured AI responses
- RevenueCat for subscription management
- Render for backend deployment
The backend layer ensures:
- Secure API key handling
- Controlled prompt engineering
- Conversation history management
- Token optimization for stable responses
Special care was taken to handle AI token limits and prevent truncated responses by limiting conversation history and increasing output capacity.
Challenges
One of the most significant technical challenges was managing AI output limits. Long coaching responses were sometimes cut mid-sentence due to token constraints.
Solving this required:
- Optimizing prompt structure
- Limiting historical context dynamically
- Increasing output token allocation
- Testing edge cases in longer sessions
Another challenge was product design.
The interface needed to feel calm and distraction-free while guiding users into structured thinking. Achieving minimalism without losing clarity required multiple UI iterations.
others challenge was:
- Forcing Gemini to strictly follow a 5-step structure without markdown leaks or vague answers (many prompt iterations!)
- Handling the first message with pre-filled context without restarting the flow
- Making RevenueCat simulation feel real in demo without real purchases
- Polish UI last-minute to go from "too minimal" to "premium minimal"
- Submitting without TestFlight public due to Apple Developer constraints – but full flow shown in video
What I Learned
This project deepened my understanding of:
- Prompt engineering for structured AI behavior
- Backend-AI communication patterns
- Token management strategies
- Designing AI experiences that prioritize intention over novelty
- Integrating subscription systems into a SwiftUI architecture
Most importantly, I learned that great AI products are not about intelligence alone — they are about framing.
Vision
ClarityCoach is designed as a foundation for a broader structured decision-making platform.
Future iterations will include:
- Long-term goal tracking
- Session memory and summaries
- Personalized coaching insights
- Progress analytics
- Expanded premium features
AI is powerful. But structured AI is transformational.
ClarityCoach is an experiment in building that transformation.
Log in or sign up for Devpost to join the conversation.