Member of Technical Staff · Jun 2025 – PresentIntern · May 2024 – Jun 2025
An agent that can query, remember, and find.
I built first versions of the capabilities that let agents work with structured data, carry context across conversations, and find answers in messy files.
Query data
A natural-language-to-SQL layer turns questions into structured queries, including complex filters and aggregations.
Remember context
Persistent memory carries useful context across conversations, with curation, nudges, and knowledge-graph synchronization.
Find answers
Improved extraction from PDFs, images, and tables raised answer correctness from 37% to 48%.
Also from 0 → 1: internal model access and knowledge integration for semantic search.



