Oracle AI

Built an AI-powered RAG assistant for intelligent querying over documents and GitHub repositories using hybrid search that combines semantic retrieval with BM25 and Cohere reranking.

Implemented repository ingestion, OCR-powered document processing, contextual retrieval, and grounded response generation pipelines for reliable answers over user-provided sources.

Designed the full-stack architecture with vector databases, backend APIs, conversational workflows, and persistent chat history management.

Integrated Voyage AI embeddings, Pinecone vector search, Groq LLM inference, Tesseract OCR, and AssemblyAI-powered speech-to-text workflows.