← Back to projects
Jan – Feb 2026 · 2 months
STRATIO, INC.
Cleaned and de-identified over 3,000 customer inquiry emails and consolidated conversation-level data for a RAG chatbot.
PythonBeautifulSoupPresidioCLD3Gemini API
Overview
A Silicon Valley global internship project transforming over 3,000 customer inquiry emails into retrieval data for a RAG chatbot. I implemented unstructured body cleaning and multilingual de-identification, then consolidated conversation-level datasets.
- Period: January – February 2026 · 2 months
- Team: 2 data engineers
- Role: Data engineering
Contributions
Preprocessing
Implemented mbox/MIME parsing and body cleaning by email type. Built an automated pipeline structuring 3,000+ inquiries into CSV documents for RAG retrieval.
De-identification
Routed privacy processing to Presidio or the Gemini API according to language.
Batch Recovery
Checkpoint resume and result reuse reduced re-execution time after a failure from 20m 46s to 4m 20s.