I'm a PhD student at uwdb and uwplse advised by Dan Suciu. I optimize modern data systems with advanced techniques from programming languages and databases:
[EGRAPHS'22] We tease out deep connections among e-graphs, finite-state tree automata, and version space algebra. By fusing together powerful ideas from these different perspectives, we gain orders-of-magnitude speedup and discover a new technique for automated proofs.
[PODS'22 best paper] datalogo is a query language that generalizes datalog. It can express a wide range of computation from shortest paths to network centrality. In the paper we study its convergence behavior.
[SIGMOD'22 invited to SIGMOD Record] We also developed a query optimizer for datalogo based on program synthesis.
[POPL'22] Using advanced join algorithms, we made pattern matching in a rewrite system asymptotically faster.
[MLSys'21] tensat is an optimizer for deep learning inference using equality saturation. It achieves state-of-the-art inference speed with very fast compilation.
[POPL'21 distinguished paper] egg is the rewrite engine underlying a new class of optimizers including tensat and spores (see below). It implements an efficient algorithm for equality saturation.
[OOPSLA'21 distinguished paper] We used egg to invent rewrite rules that are in turn given to egg itself for equality saturation.
[VLDB'20] spores is an optimizer for large scale linear algebra. It transforms linear algebra code through the powerful abstraction of relational algebra.
I have been very lucky to work with many brilliant undergraduate / master students:
Jonathan Leang meticuously performed many intricate experiments for spores. Jonathan now works at Snowflake, and at the same time (!!) teaches Databases at UW.
Yihong Zhang turned our idea of relational pattern matching into implementation and a publication. With this work, Yihong won 1st place in the PLDI 2021 Student Research Competition.