Frank Sifei Luan

Frank Sifei Luan

栾思飞

UC Berkeley

Hi!

I obtained my Ph.D. in computer science from the University of California, Berkeley in 2024. I was advised by Professor Ion Stoica. My dissertation is on An Extensible Architecture for Distributed Heterogeneous Processing. Before that, I worked at Facebook from 2017 to 2019. I obtained my bachelor’s degrees in computer science and statistics from the University of Chicago in 2017.

I am an instrument-rated private pilot with over 500 hours of flight experience.

Education

  • Ph.D. in Computer Science, 2024

    University of California, Berkeley

  • B.Sc. in Computer Science, 2017

    University of Chicago

  • B.A. in Statistics, 2017

    University of Chicago

Publications

September, 2023 SIGCOMM 2023
Exoshuffle: An Extensible Shuffle Architecture

An extensible shuffle architecture that offers competitive performance and scalability as well as greater flexibility than monolithic shuffle systems.
PDF DOI Talk
January, 2023 arXiv
Exoshuffle-CloudSort: The 2022 CloudSort Benchmark Winner

Winner of the 2022 CloudSort Benchmark (Indy category) for sorting 100TB data at $0.97/TB.
PDF Code Project Blog
June, 2022 SIGMOD 2022
Balsa: Learning a Query Optimizer Without Expert Demonstrations

A query optimizer built by deep reinforcement learning.
PDF Code DOI
April, 2021 NSDI 2021
Ownership: A Distributed Futures System for Fine-Grained Tasks

A decentralized object metadata ownership system for fine-grained distributed tasks.
PDF
February, 2021 IEEE Software
AI in Software Engineering at Facebook

Three productivity tools that learn patterns from software artifacts that we deployed at Facebook.

🏆 2021 IEEE Computer Society IEEE Software Magazine Best Paper Award
PDF DOI
June, 2020 VLDB 2020
NeuroCard: One Cardinality Estimator for All Tables

A join cardinality estimator that builds a single neural density estimator over an entire database.
PDF Code DOI
October, 2019 OOPSLA 2019
Aroma: Code Recommendation via Structural Code Search

A code recommendation tool for big code corpora to improve developer productivity.

🏆 2019 ACM SIGPLAN Distinguished Paper Award
PDF Code DOI Blog Talk
June, 2018 MAPL 2018
Retrieval on Source Code: A Neural Code Search

A natural language code search tool for big codebases.
PDF DOI Blog Dataset

Talks

Last updated on Jun 29, 2020
Curry On 2019: Using ML for Code Discovery at Facebook

We created two techniques that apply machine learning to code discovery problems: Neural Code Search (NCS) and Aroma.
Video
Last updated on Jun 29, 2020
F8 2019: Using Machine Learning for Developer Productivity

At Facebook, we use machine learning to discover patterns in code and build tools that improve developer productivity. In this session, you’ll learn about the tools we have built for code search, code recommendation, and automatic bug fixing.
Video

Posts

Flying Milestones
My journeys as a private pilot.
Last updated on Sep 19, 2023 3 min read Flying
Prehistory
Visit my prehistoric homepage.
Last updated on May 4, 2020 1 min read