AI-powered financial assistant with voice and chat interfaces
About the Project
Led the design and implementation of real-time voice and chat-based agentic AI solutions for Lazervault (AI-powered financial assistant). Built full-scale web/mobile apps allowing users to manage their accounts, plots, get odds and perform safe financial transactions using natural language.
Key Features: - Node.js (TypeScript) backend with LiveKit for real-time voice - OpenAI STT/TTS integration for voice interactions - Next.js with React for full-scale web/mobile applications - Retrieval-Augmented Generation (RAG) pipelines with ChromaDB - Fine-tuned LLaMA models for intelligent responses - GraphQL APIs for high-volume FinTech consumers - Azure Kubernetes Service (AKS) and Docker for orchestration - Firebase, Cloud Functions, and cron jobs - AWS SQS/SNS and API Gateway for event-based processing - Internationalized frontend with Material UI and Tailwind CSS - Strapi CMS for dynamic content and multi-language support - GitHub Actions CI/CD for automated deployments
Key Achievements
- Built RAG pipeline improving response accuracy by 45%
- Implemented real-time voice AI with 99.9% uptime
- Reduced transaction processing time by 70%
- Scaled to serve thousands of concurrent financial transactions