At the Code with Kiro Hackathon, my teammates and I set out to solve a challenge many people face but rarely talk about:
👉 How do you prepare for conversations in completely new environments?
Whether it’s joining a new workplace, attending your first college class, or navigating cultural/language barriers, communication can feel intimidating. Our solution was PersonaPrep (launching soon 🚀), an AI-powered personality coach that lets users practice conversations and build confidence in a safe, supportive environment.
This blog is about how we built it, and more importantly, how Kiro supercharged our development process.
🧩 Why an AI Personality Coach?
PersonaPrep focuses on three big problems:
- Social anxiety → Rehearsing workplace introductions, meetings, or even casual small talk.
- Adaptive learning → Tailoring practice sessions based on user needs and past conversations.
- Language barriers → Practicing common phrases in a new language or cultural context.
Instead of a static chatbot, PersonaPrep is like a personal coach that learns and guides you to drive conversations confidently in a new environment.
⚙️ Tech Stack
We wanted a setup that was robust yet hackathon-friendly, so we chose:
- Backend: Spring Boot (Java 17) – for REST APIs and WebSocket support
- Database: MongoDB Atlas – storing user sessions & conversation history
- WebSockets: Real-time interactive dialogues with the AI
- Vector Search: To recall and suggest similar past practice scenarios
- Frontend: React + Tailwind – for a clean, fast UI
🚀 How Kiro Accelerated Development
This is where Kiro really shone. Instead of spending hours setting up boilerplate or debating project structure, we leaned heavily on Kiro’s AI coding capabilities.
1. AI Code Generation
Kiro quickly bootstrapped core backend pieces:
- Spring Boot controllers for REST endpoints
- MongoDB repositories for user data
- WebSocket configuration for real-time chat
This gave us a solid skeleton we could extend in minutes.
2. Spec Mode
The spec mode in Kiro was a game-changer. Instead of asking vague “generate code” prompts, we described features at a high level.
For example, we wrote: “Help us with a boilerplate code for the PersonaPrep application based on the Steering Docs.”
Kiro not only outlined the tasks but also executed the code generation step by step. It felt like pair programming with a senior engineer who also kept the Jira board up to date.
3. Steering Docs
Hackathon projects often risk losing focus — adding too many features or drifting from the core idea. Kiro’s steering docs acted like a living product spec.
We defined PersonaPrep’s vision in the steering docs:
- Stay user-first (practice conversations, not generic Q&A).
- Support real-time dialogue (not batch prompts).
- Keep responses contextual (memory + vector search).
Every time we iterated, Kiro aligned its code suggestions with these docs. This saved us from rework and kept the app scoped correctly for hackathon deadlines.
4. Hooks
Kiro saved a lot of time by automating manual tasks by creating hooks. These hooks defined the tasks to be executed automatically after every defined action, thus saving all the manual effort.
🔍 Example Flow
Here’s how it works end-to-end:
- User enters scenario → e.g., “I am starting a new job in Germany, and I don’t speak fluent German.”
- AI Coach responds with tailored dialogue → “Try introducing yourself like this…”
- Conversations stored in MongoDB → allows reflection and tracking growth.
- Vector search recalls similar sessions → surfaces useful practice history.
- User continues practice in real-time via WebSockets.
📖 Lessons Learned
- Kiro as a teammate: Beyond just code, Kiro gave us structure, specs, and clarity.
- AI accelerates hackathons: With boilerplate handled, we spent time on features that mattered.
- Scope discipline matters: Steering docs ensured we built something shippable, not just “cool demos.”
💬 Final Thoughts
PersonaPrep wasn’t just a hackathon project — it showed us the power of AI tools like Kiro to bridge the gap between ideas and production-ready code.
With Kiro’s spec mode and steering docs, we shipped a full-stack AI application in record time.
If you’re building at hackathons (or even production projects), Kiro isn’t just an AI assistant — it’s a project accelerator.
Top comments (0)