Talk2.AI
Retrieval-Augmented persona simulation powered by LLMs with long-term conversational memory.
- Category
- Generative AI
- Status
- Production
- Stack
- Python, LangChain, Groq +4
Problem
Traditional chatbots struggle to maintain personality consistency and contextual memory across multiple conversations.
Architecture
Documents are embedded using SentenceTransformers, stored in ChromaDB, retrieved through LangChain, and passed to Groq's LLaMA model with MongoDB-powered conversational memory.
Key features
- Persona-based conversations
- Retrieval-Augmented Generation
- Long-term conversational memory
- Voice synthesis with ElevenLabs
- Custom persona creation from user documents
My role
Designed and implemented the complete RAG pipeline, vector database integration, conversational memory, and backend workflow.
Impact
Showcases advanced conversational AI techniques including RAG, vector search, persona simulation, and persistent memory.
Stack
A decentralized, zero-commission tutor marketplace that connects students and parents directly with local and online tutors.