• Foundation Model: MuRIL (- tokerizer aru herne ) TODO: look for finance pre-trained models
• SQuAD ma fine tune (combining multiple datasets ?)

Performance Evaluation

Benchmarking dataset: - XQuAD (hindi)
- gold standard dataset banaune ? - textual similarity - every domain bata few datasets banaune ~300 examples for benchmark

TODO

• Translation garidai xa
• Fix domain(national economics) , collect data
  • public financial dataset -> SQuAD
• RAG: vector store redis
  • langchain
• Build an application

RAG: QA model ->