Query engine development
Extending Velox, Presto, and Spark with semantic query capabilities, alongside interfaces for analysts and engineers.
Enterprise Data Lakes / 2026–present
Semantic Query Processing Engine (SQPE)
Bringing LLMs into enterprise query processing.
I lead the design and implementation of a Semantic Query Processing Engine for enterprise data lakes. The project extends established query engines with LLM-based operators and studies when their results can be reproduced or reused. Development is ongoing.
Reliability and reuse
Studying variation between repeated query executions and conditions for safely reusing LLM-generated labels across queries.