Experience
Agentic AI Engineer Intern
Alimento Agro Foods, May 2026 to June 2026
Checked distributor promotional claims against bills that arrived handwritten, scanned, or digital.
- Gemini
- FastAPI
- PostgreSQL
- entity matching
Document intelligence
- Built a Gemini workflow that classifies bills and extracts structured line items from mixed formats.
- Added field level confidence scores so a reliable extraction can be separated from an uncertain one.
Claim reconciliation
- Built a multi stage check of product and bill data against the promotional claim.
- Combined exact product matching, fuzzy name similarity, and price checks, because product descriptions are not written the same way twice.
- Routed ambiguous and conflicting matches to a person, instead of treating every automatic match as final.
Backend
- FastAPI with asynchronous Python, and PostgreSQL for claim history.
- Checked new bills against earlier submissions, while unresolved historical claims stayed intact.
This is a public summary. Internal documents, production data, private business rules, and company code are not included.