Kaseddie AI

An autonomous AI trading agent that acts as a "force multiplier" for solo investors, analyzing crypto markets and executing strategies 24/7.

Comment

💡 The Inspiration

Cryptocurrency markets operate 24/7, but human founders need sleep. I wanted to build a system that wasn't just a passive dashboard, but an active partner—an autonomous agent that could monitor price action, analyze sentiment using Generative AI, and verbally communicate strategies to me in real-time.

👻 What it does

Kaseddie AI is a "Frankenstein" creation, stitching together the web's most powerful APIs into a cohesive trading platform for solo founders.

  • The Brain: Powered by Google Vertex AI (Gemini 1.5 Flash), it analyzes real-time news to generate market sentiment and trading advice.
  • The Voice: Using Google Cloud Text-to-Speech (Journey Voice), the agent speaks its actions aloud, confirming trades ("Buying BTC") and explaining its reasoning naturally.
  • The Eyes: It ingests real-time market data via the Binance API to visualize price movements via the "Crypto Pulse" ticker.
  • The Nervous System: User identity is secured via WorkOS (SSO & KYC Verification), while financial operations are handled via Stripe.
  • The Hands: A custom Trading Engine executes orders based on 8 distinct algorithmic strategies (e.g., Momentum, Mean Reversion) and automatically calculates Stop Loss/Take Profit levels to manage risk.

🏗️ How I built it (The Kiro Way)

This project was architected entirely within the Kiro IDE, utilizing its AI-native workflow to move from concept to cloud deployment in days.

  1. Spec-Driven Development (Backend): I used Kiro's Spec mode to define the complex backend logic. I wrote a requirements document for the TradingEngine and UserVault, and Kiro scaffolded the Node.js/Express services, ensuring a clean separation between the AI logic (aiService.js) and the execution logic (trading.js).
  2. Vibe Coding (Frontend): I used Kiro's Vibe Coding to build the React interface. I prompted for a "Spooky Dark Mode" with glassmorphism effects, and Kiro generated the Tailwind CSS components instantly. It was also crucial for debugging complex CORS issues between the frontend and backend.
  3. Cloud stitching: The app is deployed as a distributed system, with the Node.js backend on Render and the React frontend on Netlify, communicating via secure API endpoints.

🧠 Challenges I ran into

The biggest challenge was the "Chain Reaction" of latency. Because the system stitches together news fetching, AI analysis, and voice generation, requests sometimes took 30+ seconds. This caused frontend timeouts and server crashes. Solution: I implemented a robust "Fail-Safe" architecture. The backend now uses AbortControllers and fallback data mechanisms to ensure the UI never crashes, even if an external API (like Binance) blocks the request due to geo-restrictions.

🏆 Accomplishments that I'm proud of

I am proud of migrating the voice engine from a limited tier to Google Cloud Text-to-Speech, giving the agent a professional, limitless voice. I successfully integrated WorkOS for enterprise-grade authentication, making the app feel like a legitimate fintech product from Day 1.

🚀 What's next for Kaseddie AI

The next phase is moving from "Paper Trading" (database simulation) to mainnet execution, connecting the strategy engine directly to decentralized exchanges (DEXs) via smart contracts.# Inspiration

Built With

Share this project:

Updates