AI Algorithm Intern — Document Processing for RAG-based System
Document
Data CollectionWorked on financial document processing for a RAG-based credit report generation system. The project involved collecting and processing annual reports, financial statements, and company-related documents from listed enterprises. I used OCR-based document parsing to extract information from complex layouts, cleaned and structured the extracted text, and prepared high-quality data sources for downstream retrieval and LLM-based report generation. The work included document collection, OCR result verification, noisy text filtering, table and paragraph organization, entity-related information extraction, and preparation of retrieval-ready data for Elasticsearch and GraphRAG pipelines. Quality measures included manual spot checks, consistency verification across financial years, removal of duplicated or low-quality text, and validation of retrieved evidence before it was used as context for LLM generation.