RiskXpertAI

Project Duration: Since March 2025 (Ongoing)

An AI-powered chatbot service for predicting financial indicators

• Tech Stack: FastAPI, Docker, GitHub Actions, MongoDB, Redis, AWS, Kakao Login API
• Team (2 members): Myself (Full-stack & DevOps), AI modeling teammate (Prediction models)
• GitHub: RiskXpertAI
• Live Service: RiskXpertAI

📌 Motivation
- Build a personalized prediction and analytics platform to support investment decision-making in volatile markets
- Utilize open-source AI models and real-time data processing to deliver fast and accurate insights

📘 Key Features
- Streaming-based response delivery
- Optimized Redis caching system
- Chat history stored in MongoDB
- Kakao OAuth integration for secure login
- Transformer-based time series forecasting
- GPT-3.5-powered summarization and scenario generation

🙋‍♂️ My Role & Contributions
- Designed entire system architecture and deployed infrastructure
- Developed FastAPI backend and integrated model inference logic
- Implemented streaming response handling and cache optimization
- Designed Redis caching for reduced latency and higher throughput
- Integrated Slack notifications for critical error monitoring

🐳 Deployment & Infrastructure
- Currently deployed via `docker-compose` on AWS EC2 manually
- Planning to implement CI/CD with GitHub Actions for zero-downtime deployment
- Utilizes MongoDB Atlas, Redis Labs for managed services
- Optimized for lightweight traffic and cost-efficient operation