I am an undergraduate at CMU School of Computer Science with interests in systems and machine learning.

I’m keen on exploring the nature of machine learning and scaling it to build large systems that generate exponentially more impact. These include but are not limited to recommender systems, DL frameworks and infrastructures, MLOps systems, and federated learning.

Currently

I am working with Rashmi K. Vinayak and Jack Kosaian in TheSys research group as part of the Parallel Data Lab on efficient fault tolerance strategy for embedding tables in distributed deep learning training. We tailor it to work specifically well in modern industrial recommender systems.

My journey with MLSys started from my work at ByteDance. I contributed to the graph embedding models and infrastructure in Douyin and TikTok, now serving 600M+ global users. It did not stop there. This year at Uber, I owned, designed, developed, and launched the Recommended Dishes carousel on the Uber Eats home feed, satisfying the cravings of 90M+ eaters around the world.

As some side fun, I’m incubating an open-source project Alpaca-Hub as an attempt to provide better versioning in ML workflows. Another project that concerns distributed DL training in Golang (temporarily named Groot) is in our pipeline too. Shoot me an email if you are interested.

I will be continuing my education at CMU as a fifth-year master in machine learning. At the meantime, I am seeking summer 2022 internships and/or full-time opportunities beyond.

Contact

You can email me at tianyuz2/@/andrew.cmu.edu (without the slashes of course). On GitHub, I am johnzhang1999, although I don’t post much code on there. We could also connect via LinkedIn at johnzhangty.

My pronouns are he/him/his. I can speak English and Mandarin.