RAG & Active Learning Pipeline for Information Extraction
- An Efficient Active Learning & RAG Pipeline to Extract Information from Large Banking Documents (20-60 pages).
- Developed a RAG & Active Learning Pipeline to perform Information Retrieval with > 95% accuracy using techniques such as Document Chunking, VectorDB Inferences & Rerankers.
- Extracted information from the retrieved chunks with Accuracy of > 80% using the quantized & distilled LLMs.
- Deployed the pipeline integrated with in-house Information Extraction Pipeline to ensure Data Privacy and Security.
