← Back to Projects Video RAG Pipeline
Retrieve and answer questions over video content using multimodal indexing + LLM synthesis.
Categories
GenAINLP
Tech Used
PythonVideo data processingEmbeddingsVector DBPostgreSQLRAGLLMFastAPIDocker
Problem
Users need accurate Q&A over video content without manually scanning long recordings.
Approach
- Extracted transcripts + key segments
- Indexed chunks with embeddings for retrieval
- Generated grounded answers with references to segments
Results
- Searchable video knowledge base
- Grounded answers over long videos