My Hunter × Hunter line art: Gon and Killua looking toward a slowly swirling galaxy.

Hey, I'm Peter.

I'm a Quantitative Developer at Citadel. I like music, basketball, and travel (not in basketball though). Work-wise, I like to consider myself a generalist, but I most enjoy working on impactful and performant systems with a research-like focus. Despite working in NYC, I still have a fondness for being in the middle of nowhere, having previously obtained a B.S. in Computer Science from Cornell.

Previously, I was involved with cloud / distributed computing research with Cornell's Cascade and Derecho groups, under the supervision of Ken Birman. We're working on a distributed framework for applications that need strong consistency and availability at low latency.

Please feel free to reach out if you'd like to chat!

Portrait of Peter

Previous Work

Image of Cascade.
Cascade

Cascade is a C++17 cloud application framework powered by optimized RDMA data paths. It provides a key/value API for data manipulation in distributed memory and persistent storage. Besides the API, Cascade allows injecting logic on the data paths for low-latency application.

c++ c# python linux
Image of the Distributed Texas Hold 'Em poker interface.
Distributed Texas Hold 'Em Poker

To ease the isolation of the pandemic back in Spring 2020, I made this multiplayer web game which implements a classic poker game to play with friends and family. Built with socket.io, Node.js, and Express. Link to Website

websockets node.js express
AllReduce diagram: four processes contribute values and each receives the combined result.
DCCL AllReduce

Along with Weijia Song, DCCL builds upon the Derecho framework to run shard-oriented communication primitives and implement the CCL API. This can run on data in host memory, or data in GPU memory, or take inputs from one source and leave outputs in the other.Implementation · Benchmarks

c++ RDMA MPI distributed systems