← All work

Industry work

Financial question answering system

Engineering a financial chatbot with named-entity recognition, table extraction and automated answer evaluation.

2023Role period

The work

Built an automatic evaluation framework for a financial chatbot, reducing required human evaluation time by 67%. Trained Flan-T5 XXL for financial NER and built table-extraction techniques; reported QA accuracy exceeded 90% over 401K documents.

Context

Financial question answering, document extraction and automated answer evaluation.

Scope

NLP / AI Data Science Intern at Fidelity Investments; financial chatbot evaluation, named-entity recognition and table extraction.

Confidentiality

Client source code and implementation details are not publicly shared.