Liana Patel

I'm a fifth year CS PhD student at Stanford University, where I'm very fortunate to be advised by Matei Zaharia and Carlos Guestrin. Prior to my PhD, I had the honor of working with Professor Boon Thau Loo at UPenn.

My research interest are in building scalable and efficient systems that support knowledge-intensive AI applications, which leverage large amounts of data.

Email  /  Scholar  /  Twitter  /  Github

Selected Publications

LOTUS is a query engine for LLM-powered data processing over. LOTUS introduces the semantic operator model, a declarative programming model for AI-based data transformations. LOTUS allows programmers to write state-of-the-art AI pipelines in a few lines of code for a wide array of applications, including fact-checking, biomedical extraction, complex search and ranking, and research analysis.

DeepScholar-Bench is a live benchmark and holistic, automated evaluation framework to evaluate generative research synthesis. We also develop DeepScholar-base, a reference pipeline implemented efficiently using the LOTUS API. Overall, we find that DeepScholar-bench remains far from saturated.

Table-Augmented Generation (TAG) is a unified and general-purpose paradigm for answering natural language questions over databases. TAG generalizes and outperforms prior methods, such as RAG and Text2SQL on a benchmark of complex NL questions.

ACORN is an index for state-of-the-art search over vector embeddings and structured data. ACORN introduces predicate-agnostic index construction and search methods, which allows it to serve a wide range of queries with arbitrary predicates, while also outperforming prior methods.

Compass is a semantic search system over encrypted data that offers high accuracy, comparable to state-of-the-art plaintext search algorithms, while protecting data, queries and search results from a fully compromised server.

Website template from Jon Barron