Multi Modal Agent using Deepgram and Groq LPU's and Sentence Transformers for Vector Embeddings, ChromaDB for persistent vector db storage and BM25 with ColBERT reranking for hybrid search for RAG context management

Real-time AI voice conversation system experiments. Evolved from a one-to-one robot conversation system to a multi-user, human-level call and web-based platform with emotion and gender detection, dynamic VAD, and many engineering tricks with the goal of making AI conversation as close to human as possible.

Web platform for operators to configure, test, and monitor voice agents before production launch. Supports real-time text and voice interaction, session history, and per-stage latency monitoring.